Topic
#ai-generated
Loot, blog posts and adjacent themes connected to this topic. Follow the tag to keep it in your orbit.
73Shown loot
60Shown articles
8Linked neighbor tags
Topic paths
If you want to go deeper, the adjacent tags are the fastest way to compare and branch into related workflows.
Loot
More from this topic
Text: AI-generated
AI-generated · Automatically published by LinkLoot. Use ML/QSAR models to assess absorption, distribution, metabolism, excretion and toxicity for small molecules before expensive experimental testing. AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. What it does This OpenClaw skill predicts key pharmacokinetic and safety properties for drug-like compounds. It covers Absorption (Caco-2 permeability, HIA), Distribution (volume of distribution, plasma protein binding, blood-brain barrier penetration), Metabolism (cytochrome P450 interactions), Excretion (clearance, half-life) and Toxicity alerts (hERG inhibition, DILI risk, Ames mutagenicity). The tool outputs structured profiles with confidence scores and actionable recommendations, helping researchers filter large compound libraries against Lipinski, Veber and other drug-likeness rules. Who should use it Medicinal chemists, computational biologists and drug discovery teams who need rapid in-silico triage of candidate molecules. It is particularly useful for prioritizing synthesis efforts by flagging structural liabilities—such as potential hepatotoxicity or poor oral bioavailability—before committing resources to wet-lab assays. Setup surface The skill requires Python and the RDKit library (pip install rdkit). Optional advanced modeling capabilities depend on deepchem and admet-x. Users interact via command-line interfaces provided by the skill scripts, such as /admet "SMILESstring" --full for single-molecule analysis or batch processing using .sdf files. Documentation references specific model types (ML/QSAR) but does not specify which underlying algorithms are active without further inspection. Runner test plan Future validation will include static checks for dependency integrity and injection vulnerabilities. Testing will verify that script arguments parse correctly and that output formats match documented Markdown structures. Sandbox execution will confirm that predictions generate without runtime errors for known SMILES strings. No human review or live experimental comparison has occurred; all claims rely solely on the provided technical documentation. Risk notes Predictions are approximate and carry inherent uncertainty. Novel chemical scaffolds often fall outside model applicability domains, leading to unreliable results. The tool explicitly warns against over-reliance: experimental validation remains mandatory for critical decisions. In vitro-in vivo gaps mean computer-predicted clearance or toxicity may not translate directly to living systems. Species differences also affect accuracy, as many models train on animal data while targeting human outcomes. Low-confidence flags require immediate manual investigation rather than automated rejection. Source links Original skill documentation Skill listing
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
Text: AI-generated
AI-generated · Automatically published by LinkLoot. A tmux plugin that reads the state of 22 coding agents across your panes and shows blocked, working, done or idle in one tree. AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. What it does tmux-agent-wizard (taw) puts a state dashboard inside the tmux you already run. It watches every pane with herdr's detection engine and labels each agent as blocked, working, done or idle. Press prefix + A for one tree of sessions, panes and worktrees; Tab hops to the next agent that needs attention, Enter lands you in its session. If an agent blocks or finishes while you're elsewhere, a toast names it and prefix + O jumps there. It also folds in two familiar jobs: project-to-session picking from tmux-session-wizard, and save/restore of sessions, layouts and folders so agents come back into their conversations after a reboot. Your existing tmux.conf, keys, plugins and status line stay untouched. There is no central daemon — just a small watcher per pane, about 9 MB each — and no account. Who it helps Developers who juggle several AI coding agents at once: Claude Code in one window, Codex in another, OpenCode in a worktree. The pain it targets is real — one agent stuck on a permission prompt for twenty minutes while you flick through windows to notice. taw surfaces exactly which pane is waiting, including over ssh via terminal notifications or system toasts on macOS and Linux. Getting started Requirements: tmux 3.2 or newer on macOS or Linux, plus TPM. Add set -g @plugin '27medkamal/tmux-agent-wizard' to /.tmux.conf, press prefix + I, and TPM installs it. On first load taw downloads its binary from GitHub releases, checked against the release's SHA-256 checksum. Then prefix + A lists the agents already running. Optional extras: run taw integration install claude (or codex, opencode, cursor) once per agent so restore returns it to the same conversation; symlink taw into /.local/bin if you want it outside tmux. Without TPM, from source, or with the download disabled, see the project's install page. Limits and costs The detection list covers 22 agents via herdr's rules, so unusual or brand-new tools may not be recognized until the engine updates. Restore requires per-agent integration setup. Each pane runs a 9 MB watcher, which adds up on wide setups. The project is MIT-licensed and free; pricing beyond that is not stated. No security audit or human review is claimed by the documentation — the checksum verification is the only integrity step described. Source links Official repository Project documentation Project license
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
Text: AI-generated
AI-generated · Automatically published by LinkLoot. CI Lifeboat automatically falls back to your self-hosted runner only for verified quota refusals, preserving your existing CI gates. AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. What it does CI Lifeboat is a fallback controller for GitHub Actions. It monitors your trusted commits and detects explicit billing or quota refusals from GitHub’s hosted infrastructure. When that specific condition occurs, it dispatches a restricted QA workflow to an official self-hosted runner you already manage. The tool operates on an evidence-first principle. It ignores ordinary test failures, queued jobs, or network blips. The fallback triggers only after verifying the current protected-branch SHA and confirming a documented refusal. To maintain security boundaries, the self-hosted workflow rechecks this evidence before checking out any code. CI Lifeboat uses two outbound connections via the GitHub API; it never contacts your runner directly, requiring no inbound ports or tunnels. Who it helps This tool targets engineering teams using GitHub Actions who need continuous quality assurance regardless of external billing limits. It suits organizations with strict branch protection rules where a temporary loss of hosted compute shouldn't halt validation. Because it preserves existing hosted CI as the default, it fits teams unwilling to migrate their entire pipeline but wanting resilience against quota exhaustion. It requires Python 3.11+, the GitHub CLI (gh), and an independently registered, restricted GitHub Actions runner. It is not a general CI service or a substitute for production release gates. Getting started Installation involves cloning the repository and copying the example TOML configuration to /.config/ci-lifeboat/config.toml. You must edit this file to replace examples with approved private repositories. Public repositories are denied by default for safety. Before enabling active jobs, follow the step-by-step installation guide to install the restricted workflow and configure runner-group access. Dry-run mode is available to verify behavior without dispatching actual workflows. Optional systemd timers can automate the polling process. The project provides agent-specific installation rules (AGENTS.md) to help coding agents execute safe setups, but human approval is required before changing GitHub settings or registering runners. Limits and costs CI Lifeboat is currently in early preview (v0.1.0). It handles only confirmed GitHub billing/quota refusals; it does not address other CI failures. While the software is Apache-2.0 licensed, hosting costs depend on your self-hosted runner infrastructure. The tool skips previously dispatched fallbacks for the same commit to prevent redundant runs. Source links Official repository Project documentation
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
Text: AI-generated
AI-generated · Automatically published by LinkLoot. SHOBR automates the tedious parts of job hunting—scraping, resume tailoring, and application prep—by driving your actual authenticated browser. No headless bots, no credential extraction. AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. What it does SHOBR (Simulate Human Occupational-Bureaucratic Rituals) is a UNIX-style job search automation tool that prepares applications while keeping you in control. It uses beachpatrol to drive your existing daily-driver browser for scraping jobs with your real authenticated sessions, then leverages minimal LLM calls to tailor resumes from Markdown into ATS-readable PDFs via Groff and Pandoc. The workflow follows a human-in-the-loop model: SHOBR discovers jobs, screens them, writes drafts, and builds deliverables—but never auto-submits. You approve each application step manually. All data lives in append-only JSONL event logs, so you can interrupt pipelines or recompute decisions with new rules anytime. Who it helps Job seekers who want to automate repetitive tasks (finding listings, tailoring CVs) without resorting to "spray and pray" mass applications or risky headless browsers that trigger detection. Particularly useful for those targeting platforms like LinkedIn where authenticated access matters and bot detection is aggressive. Developers comfortable with Python and CLI tools will appreciate the modular design. The project avoids demanding an agent framework—it's lean enough (3500 LOC) to extend yourself if needed. Getting started Prerequisites: Install beachpatrol first (the hard dependency). See its README for setup. Python 3.14+ Installation: The toolchain requires Groff and Pandoc for PDF compilation. Ensure these are available on your system. LLM configuration: SHOBR uses any-llm abstraction, supporting OpenAI, Anthropic, local models, or even hijacking subscriptions via faaah. Configure your preferred provider before running. Limits and costs No auto-submit: Final application clicks remain manual by design. LLM costs: Resume tuning needs some LLM usage; budget for API calls if using cloud providers. Browser dependency: Requires beachpatrol, which itself needs proper setup as your daily driver. MIT licensed: Free to use, modify, distribute. Hosting/execution costs depend on your infrastructure choices. Early-stage tool: Built with stdlib unittest, full E2E TDD, but expect rough edges typical of personal projects. Source links Official repository Project documentation License
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
Text: AI-generated
AI-generated · Automatically published by LinkLoot. Run scheduled tasks and background workers in Django apps, scrapers, and CLIs using only SQLite—no broker, no daemon, zero dependencies. AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. What it does tinyjobs brings Celery-style task queuing to Python projects that don't need the infrastructure overhead. Instead of requiring a Redis or RabbitMQ instance before running a single job, it uses a plain SQLite file as its queue. You define tasks with decorators supporting retries, timeouts, and async execution, then enqueue them from anywhere—including synchronous code. The library handles the lifecycle: queued → running → success or failure. When you need current status, call job.refresh() explicitly; attribute access never triggers I/O silently. Results are off by default since most jobs run for side effects, avoiding write amplification on hot paths. Terminal jobs get pruned automatically. Successes stay for a week, failures for thirty days (because those are what you debug). Multiple workers coordinate pruning through a single conditional UPDATE on a marker row—forty workers still means one sweep per interval, no leader election needed. Who it helps This targets developers building Django email senders, nightly scrapers, or CLI tools that offload slow work. If your project never needed distributed task queues but wants reliable scheduling, tinyjobs fits. The documentation notes that when you outgrow SQLite, switching the URL to Redis leaves your task code unchanged. Getting started Install via pip, then initialize with a SQLite path: Define tasks with decorators like @app.task(maxattempts=5, timeout=300). Enqueue returns a snapshot of the database row—not a live handle—so the worker process updates state independently. Use job.wait(timeout=30) to block until completion, or await job.awaited(timeout=30) from async contexts. CLI commands manage retention policies directly: tinyjobs worker --retain-failed 90d or tinyjobs purge --before 30d --dry-run. Limits and costs Pruning doesn't shrink the SQLite file; freed pages get reused internally, so databases plateau rather than contract. Reclaiming filesystem space requires manual VACUUM, which demands exclusive access. MIT-licensed open source. No pricing information provided in the packet. Source links Official repository Project documentation License
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
Text: AI-generated
AI-generated · Automatically published by LinkLoot. Discover inventory, launch display and video buys, and track ROI through an agent skill that speaks plain English instead of dashboard clicks. AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. What it does The AdCP Advertising skill connects OpenClaw agents to the Ad Context Protocol, letting marketing teams run programmatic campaigns without learning every ad platform's interface. You describe what you want in natural language—"find premium video placements for luxury brands" or "create a $10k display campaign targeting tech professionals in California"—and the skill handles discovery, creation, creative uploads, and performance tracking. It supports eight core operations: capability discovery, product search, format listing, media buy creation and updates, creative sync and retrieval, and delivery metrics. Pricing models include CPM, auction-based bidding, CPCV for video completions, flat fees, and cost-per-point. Targeting overlays cover geography, demographics, behavioral interests, and contextual keywords. Who should use it Media buyers managing multi-publisher spend, agencies automating client reporting, e-commerce brands launching retargeting campaigns, and startups running lean marketing budgets. The skill reduces manual setup time by replacing complex dashboards with conversational commands. If your team already uses Facebook or Google ads but struggles with cross-channel optimization, this abstracts those platforms behind one interface. Setup surface MIT-licensed and open-source via GitHub. A public test agent at test-agent.adcontextprotocol.org requires no authentication for development—you can prototype workflows immediately. Production use needs valid auth tokens per publisher agreements; the documentation notes real campaigns may take minutes to days due to human approval steps. No hosting costs are mentioned, but actual ad spend remains separate from the tool itself. Runner test plan Static checks will verify JSON schema compliance for brand manifests and package structures. Dependency scanning covers the @adcp/client npm import shown in examples. Injection tests probe how the skill parses free-text briefs into structured queries. Sandbox execution against the test agent validates response handling for pending, completed, and failed status fields. Screenshots capture campaign creation flows. Residual risks include token expiration during long-running approvals and mismatched pricing option IDs between publishers. Risk notes AdCP is explicitly asynchronous—expect delays on campaign launches. The docs warn against assuming real-time responses; polling logic is required. Test agent credentials are public, so never reuse them in production. Budget errors (negative values, missing fields) return validation failures rather than silent drops. Publisher-specific approval workflows mean some campaigns stall until humans sign off. The skill inherits whatever targeting precision each connected ad network provides, which varies widely. Source links Original skill documentation Skill listing Awesome OpenClaw Skills ClawHub registry
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
Text: AI-generated
AI-generated · Automatically published by LinkLoot. Upload a project export to get a quality report in 10 seconds, or run locally with zero network calls. Detects uncalled timers, dead variables, and visualization binding errors across 500+ environments. AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. What it does PLC Lint performs static analysis on CODESYS and PLCopenXML projects to identify logic errors that pass compilation but fail in production. While standard compilers validate syntax, they ignore semantic issues like uncalled timers, truncated setpoints, or division by variables that may be zero. The tool parses project exports (.xml or .export) to detect these patterns. It supports two input formats. For standard PLCopenXML exports (tc60200), it analyzes program code from CODESYS 3.5, TwinCAT3, WAGO e!COCKPIT, Schneider EcoStruxure Machine Expert, OWEN, Weintek, and other environments built on the CODESYS runtime. For local CODESYS ExportFiles (.export), it additionally inspects device trees and visualizations. This allows it to catch hardware-specific issues such as sensor types cleared on non-digitized channels, outputs bound without signal selection, modules excluded from polling, and mismatched variable captions in UI elements. The engine uses streaming parsing to handle large files efficiently; a 29 MB export consumes approximately 47 MB of memory and processes in about two seconds. It distinguishes between warnings and critical findings based on format capabilities—for instance, "nothing writes this alarm" is raised to critical severity only when visualization bindings are available to confirm the lack of usage. Who it helps Automation engineers and PLC programmers working with CODESYS-based systems benefit from catching integration bugs before deployment. Teams using multiple vendor platforms (Beckhoff, WAGO, Schneider) can use a single rule set since all rely on the underlying CODESYS runtime structure. It is particularly useful for validating complex projects where manual review of hundreds of POUs and global variables is impractical. Getting started Two methods are documented. First, use the hosted service at plc-linter.com by uploading a project export; no signup is required for basic reports. Second, run locally: Windows users download plc-lint.exe from GitHub releases and drag-and-drop XML or export files. Cross-platform users run python cli.py project.xml. Local execution explicitly avoids network calls, enforced by internal tests replacing socket functionality. A live sample report is available at plc-linter.com/demo. Limits and costs The source describes the tool as free. The MIT license permits open-source use. However, specific pricing tiers for advanced hosted features (sharing, history, certificates) are not detailed in the provided evidence. Certain checks do not apply to .export files due to structural differences; watchdog and duplicate-Modbus-address rules remain silent for these inputs rather than guessing. Analysis depends on export completeness—missing task configurations limit some diagnostics. Source links Official repository Project documentation
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
Text: AI-generated
AI-generated · Automatically published by LinkLoot. A NeoForge mod that replaces procedural terrain with actual satellite and survey data, so mountains, rivers and biomes match their real-world counterparts. AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. What it does Earth Platform is a Minecraft mod that generates the entire planet at a 1:1 scale, where one block equals one meter. Unlike standard Minecraft worlds that rely on procedural noise, this project uses open satellite and survey data to define elevation, land cover, and climate. The result is a world where Everest stands 8,848 blocks high, the Grand Canyon displays its actual rock layers, and forests or deserts appear exactly where they exist on Earth. The mod extends the vertical build limit significantly, allowing players to climb, mine, and build on real-world peaks. It also integrates vanilla Minecraft mechanics; caves, ores, villages, and structures generate normally within the real-world terrain. Players can see distant terrain up to approximately 94 kilometers away using textured voxels, eliminating the typical fog wall and providing a clear view of far-off mountain ranges. Who it helps This tool is designed for Minecraft players interested in geography, education, or large-scale architectural projects. It serves educators who want to visualize topography and survival players looking for a unique, realistic challenge. Because it supports dedicated servers, it is also useful for communities wanting to share a persistent, accurate map of the world. Getting started To use Earth Platform, you need Minecraft version 1.21.1 and NeoForge 21.1.256 or newer. The process involves installing the appropriate NeoForge loader and then downloading the latest Earth Platform jar file from the project's releases page. Once installed, you can launch the game and select an Earth world. In creative mode, pressing 'N' opens a travel map to fly to any location instantly. In survival, commands like /earth visit allow teleportation, while elytra flight above Y 9,200 lets you ride the jet stream across continents. Limits and costs While the software itself is free and open-source under the LGPL-3.0 license, running it requires substantial hardware resources. Drawing terrain out to 397 kilometers demands very powerful computers. Additionally, the mod intentionally omits cities; built-up areas are replaced by surrounding wild land, though vanilla villages still generate. Users should note that while the code is open, hosting a server for many players incurs standard infrastructure costs. Source links Official repository Project documentation Project license
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
Text: AI-generated
AI-generated · Automatically published by LinkLoot. My Kids TV swaps the stock smart-TV launcher for a quiet, PIN-protected home screen that surfaces only vetted titles from apps you already own. AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. What it does Most smart-TV home screens are engineered to keep eyes on ads and autoplay. My Kids TV replaces that launcher with a single large window showing one title at a time — its name, a one-line summary, and why it's worth watching (Must-watch, Learn something, or Kids love it). Three thumbnails below preview what's next. The display rotates every 10 seconds and refreshes fully every 30, but freezes the instant a remote button is pressed so a child can browse without distraction. Titles come from curated catalogs split by age band: 4–6 (86 titles), 7–10 (314), and 11–13 (400). A title only appears if it lives in an app already installed on the TV — Netflix, Prime Video, Disney+, JioHotstar, and others — and launches straight into that app. Once opened, it drops off future suggestions. Behind a parent PIN sit controls for switching age bands, hiding specific apps from the home screen, disabling the YouTube button on supported remotes (Haier, TCL, Hisense, Sony), setting a sleep timer from 5 minutes to 4 hours, and restoring the TV's original launcher with no root required. Who it helps Families who want a dedicated kid-friendly TV interface without buying separate hardware. It runs on Google TV and Android TV as a launcher replacement, and on LG (webOS) and Samsung (Tizen) as a standalone app. Getting started Install from the project's GitHub repository. The app communicates only with TMDB for title metadata and Open-Meteo for weather; viewing history stays local on the device. No account creation is needed. Limits and costs The software is MIT-licensed and free. Hosting and distribution through each TV platform's store may impose their own requirements not covered here. The catalog size per age band is fixed at the counts above; expanding it depends on the maintainers' review pipeline, which combines best-of lists, TMDB sweeps, automated scoring via TypeSafe's Jev model, and human review of borderline picks. App-hiding disables apps rather than uninstalling them. Sleep timer range caps at four hours. Source links Official repository Project documentation MIT License
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
Text: AI-generated
AI-generated · Automatically published by LinkLoot. OpenXet lets you run a chunk-deduplicated Xet server so official hf and Git clients upload only changed file chunks to your own infrastructure. AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. What it does OpenXet provides an open-source server and client implementation for the Xet protocol, which is Hugging Face's method for storing large, versioned files using content-addressable storage (CAS) with global deduplication. Instead of uploading entire files repeatedly, only modified chunks are transferred. The project currently supports Milestone M1: openxet serve runs as a local CAS that stock hfxet (version 1.7.0+) uploads to and downloads from directly. It includes verified uploads, scoped tokens, and signed download URLs. Reconstruction performance answers 100k-term file requests within 50 ms at p99 latency on an idle node, gated in CI. The system runs on local disk with a SQLite index. Future milestones will add Hub API compatibility (hf upload, huggingfacehub, Git integration) and support for S3, GCS, Azure, and Postgres backends. Who it helps Teams managing ML models or datasets who want data sovereignty without abandoning Hugging Face tooling. If you need files to stay in your bucket while identity remains in your existing login system, OpenXet bridges that gap. The goal is unmodified operation of official tools against self-hosted infrastructure. Getting started Build the openxet binary, then follow the quickstart guide at docs/quickstart-storage-only.md. Start a server instance and test uploads/downloads with hfxet or the openxet command line. The documentation covers server configuration, CLI options, exit codes, and CAS routes. Security tests document each requirement and its enforcement mechanism. Benchmarks detail the latency gate and reconstruction metrics. Note: There is no release yet. This is development-stage software. Limits and costs Current state: Storage server works; Hub API is not implemented. Only local disk with SQLite index is supported—cloud storage backends come in later milestones. No hosted service exists yet despite the openxet.com homepage reference. Apache-2.0 licensing permits commercial use but does not imply free hosting. Performance claims are from CI-gated benchmarks, not production deployments. Source links Official repository: https://github.com/stubbi/openxet Project documentation: https://github.com/stubbi/openxet/blob/main/README.md Project license: https://github.com/stubbi/openxet/blob/main/LICENSE
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
Text: AI-generated
AI-generated · Automatically published by LinkLoot. See where your hours go with automatic desktop tracking, on-device AI daily reports, and optional Google Drive or WebDAV sync. AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. What it does Hindsight is a local-first desktop tool that automatically records app and window activity in the background. It visualizes where your time goes through stacked 24-hour timelines, daily/weekly/monthly charts, and category rankings like work or entertainment. No manual timers or project tags required. The standout feature is on-device AI: a local model can roll up your day's segments into a written report, and an AI chat interface answers questions like "How long did I spend on project X this month?" directly from your own records. Screen memory search lets you find any text that appeared on screen and jump to the screenshot and moment (screenshots and OCR are off by default). Optional multi-device sync via Google Drive or WebDAV aggregates activity data across computers for combined analysis. Who it helps People who want honest visibility into how they actually spend computer time without relying on cloud services. Developers, freelancers, and anyone juggling multiple machines will appreciate the local privacy-first approach paired with AI summaries that turn raw logs into readable insights. Getting started Download the installer from GitHub Releases. On Windows, double-click hindsightx.y.zx64-setup.exe. Note that SmartScreen will block first launch because the installer lacks an EV code-signing certificate; click "More info" then "Run anyway" to proceed. On macOS, mount hindsightx.y.zuniversal.dmg and drag Hindsight into Applications. Data stays on your machine by default. Enable screenshots, OCR, or sync features as needed. Limits and costs Hindsight is open source under GPL-3.0, but hosting costs for optional sync depend on your chosen provider (Google Drive or WebDAV). The unsigned Windows installer triggers SmartScreen warnings until proper code signing is added. AI features require either a local model setup or cloud API configuration; documentation describes capabilities but pricing for cloud options remains unspecified. Plugin marketplace and custom themes extend functionality. Source links Official repository: https://github.com/Tomotsugu-dev/Hindsight Project documentation: https://github.com/Tomotsugu-dev/Hindsight/blob/main/README.md License: https://github.com/Tomotsugu-dev/Hindsight/blob/main/LICENSE
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
Text: AI-generated
AI-generated · Automatically published by LinkLoot. TenseMem gives AI agents a single SQLite-backed memory layer that keeps one current truth per slot, tells you whether a read is fresh, and refuses to learn from its own output. AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. What it does TenseMem provides embedded, deterministic state memory for AI agents. The model proposes typed operations (add, set, correct); a rule engine decides what is actually stored. One SQLite file inside your process holds the data—no server required. Each fact lives in a named slot. A new value supersedes the old one, but history still answers “where did I live before?” Every read reports asofseq, pending, and iscurrent, so callers know whether an extraction has finished applying. Recalled context is fenced, and the extractor needs a verbatim user quote to learn, preventing self-echo contamination. Who it helps Developers building agent systems where stale or contradictory memory causes real errors. If similarity search returns both “I live in Delhi” and “I moved to Bangalore” and leaves the LLM guessing, TenseMem’s slot model gives you one authoritative answer plus an audit trail. It suits Python services, CLI tools, REST APIs, or MCP integrations that need lightweight persistence without running separate infrastructure. Getting started Install via pip with optional provider extras: tensemem[openai], tensemem[anthropic], tensemem[ollama], tensemem[litellm], or Gemini (GEMINIAPIKEY in environment). The default FakeExtractor is rule-based and needs no API key, useful for local testing. The repository includes a packaged benchmark (tensemem bench) that scores 101 conversations offline with FakeExtractor. Architecture docs map every surface—Python, CLI, REST, MCP—to source files, and the test suite locks core invariants against auto-registering unknown predicates or teaching from fenced text. Limits and costs Apache-2.0 licensed open source; hosting costs depend on your deployment. Provider keys go in process environment only. The default store is a single SQLite file with one writer thread, which may limit concurrent writes in high-throughput setups. Freshness guarantees apply within a single process; distributed deployments need external coordination not described here. Source links Official repository Project documentation Project license
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
Text: AI-generated
AI-generated · Automatically published by LinkLoot. Generate Turkish speech with inline emotion directions using a distilled speaker model that streams experts from NVMe, keeping VRAM under 15 GiB. AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. What it does This project adapts the Fish Audio Drama 3 text-to-speech model for Turkish by replacing its full speaker head with a distilled 32-expert module. The key architectural trick is streaming: instead of loading the massive 41.5 GB checkpoint into RAM or VRAM, the engine pulls expert weights directly from an NVMe drive during inference. This allows the system to run on consumer hardware with just 16 GB of VRAM. The output preserves Drama 3’s ability to interpret inline directions (like [whispering] or [excited]) written in Turkish within square brackets. Benchmarks show it runs at roughly 2.7x real-time speed for short sentences and up to 3.4x for paragraphs when using a "hot tier" cache that keeps frequently used experts in pinned RAM. Who it helps Developers and researchers working on Turkish audio generation who are constrained by memory limits. It’s particularly useful for local deployments where cloud TTS APIs are too expensive or privacy-sensitive. The project provides concrete performance metrics, including a Word Error Rate (WER) of 2.27% on the Freya-TR-Eval benchmark, making it a viable option for high-quality synthetic speech without needing enterprise-grade GPUs. Getting started Setup involves cloning the repository and installing dependencies via pip. You need to download three components: The base Fish Audio Drama 3 model from Hugging Face. The distilled Turkish speaker weights (speaker.safetensors). The expert trace file (traceexperts.pt) which informs the hot tier selection. To run inference, you use a Python script with specific flags. For example, --disk --hot 6000 enables disk streaming and caches the top 6,000 experts in RAM (requiring 25 GB total system RAM). If your machine has less than 32 GB RAM, omit the --hot flag to rely entirely on disk reads, though this drops speed to 1.2x real-time. A simple HTTP server (server.py) is also included for integration into applications. Limits and costs The software is open-source under Apache-2.0, but the underlying Fish Audio Drama 3 model may have separate licensing terms; check their official documentation before commercial use. Hardware requirements are strict: you need an NVIDIA GPU with at least 16 GB VRAM and fast NVMe storage. Peak VRAM usage hits 15.0 GiB during paragraph generation. System RAM needs vary significantly based on whether you enable the hot tier cache. An optional FP8 quantization mode exists but increases error rates slightly, so it is disabled by default. Source links Official Repository Project Documentation
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
Text: AI-generated
AI-generated · Automatically published by LinkLoot. A Go server that speaks the real Xet protocol, so the standard hf CLI works against your own endpoint — with a pull-through proxy that keeps cached repos available even when huggingface.co goes down. AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. What it does xet-server is a Go implementation of Hugging Face's Xet storage protocol, the content-defined-chunking, dedup-first layer that replaced Git LFS for large model files on the Hub. It is wire-compatible: point the real hf CLI (huggingfacehub + hfxet) at it via HFENDPOINT, and hf upload / hf download work end-to-end, producing byte-identical round-trips with genuine BLAKE3 hashing, real xorb/shard binary formats, and LZ4/byte-grouping compression. The project started as a simplified chunking demo and grew into a from-scratch, spec-driven reimplementation, verified against a live hfxet client. Several wire-format quirks were found only by capturing and replaying real client traffic (documented in docs/PROTOCOL.md). It ships two main pieces: The wire-compatible server — a CAS HTTP API (xorb upload/fetch, shard upload, file reconstruction with Range paging) plus a Hub API shim covering repo create, preupload, read/write tokens, commit, and resolve/HEAD, including independent revisions and branches per repo. xet-proxyd — a caching pull-through proxy for the real huggingface.co. This is an offline-resilience layer, not just a performance cache: any repo metadata or xorb bytes it has relayed once stay servable even after the upstream becomes unreachable, and its cache directory can be handed to a plain xetd as a permanent disconnected replacement. Storage backends are pluggable, and interactive Swagger UI docs are included. Who it helps Teams running air-gapped or bandwidth-constrained ML infrastructure who want to serve model weights through the standard toolchain without depending on Hugging Face's hosted endpoint. Also useful for anyone needing a local staging target for Xet-based workflows, or an on-prem resilience layer that survives upstream outages. Getting started Build and run the server, then set HFENDPOINT to its address and use the normal hf commands. The README walks through running the wire-compatible server, pointing the CLI at it, the Xet Data API + CLI, and xet-proxyd. A troubleshooting section covers common issues, and a "Where this diverges from real Xet" section documents known gaps. Limits and costs MIT-licensed, so source use is unrestricted, but hosting is on you — hardware, storage, and operational overhead are not covered by the license. It is a from-scratch reimplementation, so expect documented divergences from the production Xet service; review those before relying on it in critical paths. No pricing is published because there is none beyond self-hosting. Source links Official repository: https://github.com/guilt/xet-server Project documentation: https://github.com/guilt/xet-server/blob/main/README.md Project license: https://github.com/guilt/xet-server/blob/main/LICENSE.md
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
Text: AI-generated
AI-generated · Automatically published by LinkLoot. Gonic is a free-software Go implementation of the Subsonic API that streams your own collection to dozens of compatible clients, runs on modest hardware like a Raspberry Pi, and adds podcasts, multi-user support, scrobbling, and gapless jukebox playback. > AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. What it does Gonic is a self-hosted music streaming server that implements the open Subsonic server API. Because it speaks that protocol, it works with the wide ecosystem of Subsonic-compatible clients rather than locking you into one app. It indexes your existing folder tree and preserves it for browsing, while also letting you navigate by tags through taglib — covering mp3, opus, flac, ape, m4a, wav and more. Audio can be transcoded on the fly via ffmpeg and cached, and a server-side jukebox mode enables gapless playback instead of streaming. Podcasts are supported alongside music. Other documented features include multiple users each with their own playlists, top tracks and transcoding preferences; last.fm and ListenBrainz scrobbling; artist similarities and biographies pulled from the last.fm API; handling of multi-valued tags such as albumartists and genres; and an album-artist tag so compilation appearances don't clutter your artist list. A web interface handles configuration: setting up integrations, managing users and starting scans. The project reports fast scanning — roughly 10 minutes for an initial pass over about 50k tracks and around 6 seconds for incremental rescans — though real-world results depend on your library and hardware. Who it helps Anyone who owns their music files and wants remote access without renting space or handing a third party their collection. Written in Go, Gonic is described as lightweight and suitable for single-board computers like a Raspberry Pi, with ARM images available. Households benefit from per-user accounts; listeners who use apps such as Ultrasonic, DSub, Symfonium, Amperfy, Jamstash, Soundwaves or Strawberry get native client support. Getting started Point it at your collection with GONICMUSICPATH (or -music-path). Optional paths cover podcasts (GONICPODCASTPATH), playlists (GONICPLAYLISTSPATH, using numbered /.m3u layout) and a transcode cache (GONICCACHEPATH). The default login is admin/admin; change the password from the web interface once you're in. Community discussion happens on the gonic Libera.Chat channel. Limits and costs The software is GPL-3.0 licensed, but that refers to the code's freedom, not hosting: running your own server still costs electricity, storage and any bandwidth you pay for. Transcoding requires ffmpeg installed on the host. Client compatibility is based on projects the maintainers tested against, so behavior in other apps may vary. Pricing beyond a self-managed install is not specified. Source links Official repository: https://github.com/sentriz/gonic Project documentation: https://github.com/sentriz/gonic/blob/master/README.md Project license: https://github.com/sentriz/gonic/blob/master/LICENSE
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
Text: AI-generated
AI-generated · Automatically published by LinkLoot. A free, CC-licensed dataset covering every MLS regular-season and playoff match since 1996 — pre-match and post-match Elo ratings for both teams in a single CSV. AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. What it does The philo92/mls-elo repository publishes historical and current Elo ratings for Major League Soccer, covering every regular-season and playoff match from the league's inaugural 1996 season to today. Three CSV files form the dataset: results.csv lists each match with pre-match and post-match Elo ratings for both teams, full-time scores, outcomes, and neutral-venue flags; teams.csv holds each club's current rating and last update date; and homeadvantage.csv tracks the home-field advantage parameter used in the model. The methodology follows standard football Elo conventions: every team starts at 1,500, all matches carry equal weight (no bonus for playoffs or rivalries), and margin-of-victory adjustments reward larger goal differences. Only MLS league matches are included — cup competitions like the U.S. Open Cup and continental tournaments such as the CONCACAF Champions Cup are excluded. Team names are unified to each club's current name throughout, so 1990s MetroStars matches appear under New York Red Bulls, keeping a franchise's full history trackable under one identifier. Two-legged playoff rounds from 2003–2018 are combined into single aggregate results treated as neutral-site matches. Extra-time scores appear in some historical records, but going forward updates use only regulation-time results. Who it helps This is a source-of-truth dataset for anyone building MLS analytics, prediction models, or content that needs consistent power ratings across three decades of play. The project fills a gap left by ClubElo, which covers European leagues but not MLS. It also feeds oddsline.io, a football betting analytics platform run by the same solo developer, so the data is battle-tested against real modeling work. Getting started Clone or download the repository directly from GitHub. No API keys, dependencies, or build steps are required — the data ships as plain CSV files ready for pandas, R, spreadsheets, or any tool that reads delimited text. Ratings are updated manually after MLS match days on a best-effort basis, typically staying current within a few days of each round. Limits and costs The dataset is licensed under Creative Commons Attribution 4.0, which permits commercial and non-commercial reuse provided you credit the source. Hosting and access are free through the public GitHub repository. Key limitations: no cup or continental competition coverage, equal weighting across all match types, and manual updates rather than automated pipelines, so freshness depends on the maintainer's schedule. Historical extra-time scores may inflate margins compared to the regulation-only policy now in place. Source links Official repository Project documentation License (CC BY 4.0)
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
Text: AI-generated
AI-generated · Automatically published by LinkLoot. Automatically create short links, set passwords or expiry times, and submit org-scoped feature requests using the 1p.io API. AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. What it does The 1p-shortlink skill enables agents to interact with the 1p.io service for two primary purposes: creating manageable short URLs and handling internal feature requests. It provides a structured way to shorten long links before sharing them with users, offering control over link behavior through optional parameters like custom slugs, password protection, and time-to-live (TTL) settings. Beyond link management, the skill integrates with 1p.io’s feature request system. Agents can submit new ideas, browse existing proposals within their organization, vote on features, and track status updates from pending to done. This creates a feedback loop directly within the agent workflow, scoped strictly to the organization associated with the API key. Who should use it This tool is designed for developers and teams using OpenClaw who need to share clean, memorable links with end-users or stakeholders. It is particularly useful for agents that frequently output long URLs exceeding 80 characters, as the documentation suggests automatic shortening in these cases. Teams managing product roadmaps internally will also find the feature request module valuable for tracking agent-driven improvements without leaving their current environment. Setup surface To begin, you must register your instance via the POST /api/register endpoint. You have two options: Owner Registration: Provide an owneremail. This returns an API key immediately but requires verifying the email via OTP. Organization Join: Provide an organizationid. This enters a pending state until the owner approves the request via dashboard or magic link. Once registered, verify access by submitting the OTP received via email. The skill relies on Bearer token authentication for all subsequent operations. Limits apply based on verification status: unverified accounts are capped at 10 links per day, while verified accounts allow up to 1,000. Runner test plan No execution tests have been performed. A proposed validation plan includes static analysis of the SKILL.md file to ensure schema compliance. Dependency checks would verify network access to 1p.io/api. Injection testing should confirm that user-supplied URLs and slugs are sanitized according to the documented constraints (e.g., minimum 8 characters for custom slugs). Sandbox checks would simulate API calls to verify error handling for invalid tokens or exceeded limits. Screenshots of successful link creation and feature voting flows would be captured if relevant UI components are exposed. Risk notes The skill depends entirely on the availability and stability of the third-party 1p.io service. If the external API goes down, link shortening and feature management capabilities will fail. Privacy concerns arise when sending URLs to an external service; sensitive data in query parameters could be exposed. Additionally, the feature request system is organization-scoped, meaning cross-org visibility is not supported. Users must carefully manage API keys, as compromised credentials could allow unauthorized link creation or spamming of feature votes. The documentation does not specify data retention policies for shortened links or submitted requests. Source links Original skill documentation Skill listing Awesome OpenClaw Skills ClawHub registry
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
Text: AI-generated
AI-generated · Automatically published by LinkLoot. SuperCMO Skills gives Claude, Cursor, and Codex an open-source pipeline for UGC reels, ad videos, and product photography — no closed platform required. AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. What it does SuperCMO Skills is an Apache-2.0 licensed collection of agent skills that turns any capable AI coding agent into a full marketing production studio. You hand it a product photo and a brief — "make me an ad video for this energy drink" or "create a get-ready-with-me reel for this t-shirt, very genz" — and the agent plans the video, writes the script, generates AI actors, creates storyboards, produces clips, and stitches them into a campaign-ready asset. The documentation states it maintains actor and product consistency across videos of any length. Who it helps Small brands, solo marketers, and indie developers who need polished video ads without hiring a production team or paying for a closed AI marketing platform. If you already use Claude Code, Cursor, or Codex daily, this slots directly into your existing workflow rather than forcing you into another SaaS dashboard. Getting started Installation is one command: npx --yes @supercmo/skills@0.1.27 --all. It detects which agents you have and installs accordingly; for Cursor specifically, run it in your project directory with --cursor --project-dir .. After install, ask your agent to "set me up" or invoke the supercmo-setup skill. You bring your own model API keys, or sign in to pay per use through SuperCMO's hosted service at getsupercmo.ai. Every generation supports a dryrun flag so you can preview the exact request before spending anything on API calls. The hosted SuperCMO option requires zero setup — just sign in, connect social accounts (Instagram Business, Facebook Page, LinkedIn, YouTube, Reddit), and schedule posts from a built-in calendar. Limits and costs Pricing for the hosted service is not specified in the repository documentation; unknown pricing stays unknown here. Self-hosted use depends entirely on what your chosen image and video model providers charge. The skills themselves are free under Apache-2.0, but generating real video content will consume paid API credits regardless of hosting choice. No independent security audit has been performed; review the source yourself before running it against sensitive brand assets. Source links Official repository Project documentation License
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
Text: AI-generated
AI-generated · Automatically published by LinkLoot. PhotoCraft Companion adds GPU-accelerated neural tools and an overlay palette to the Rust-based PhotoCraft editor, bypassing its WebAssembly sandbox. AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. What it does PhotoCraft Companion extends the open-source PhotoCraft graphics editor with native, hardware-accelerated AI tools that run directly on your active canvas layer. Instead of exporting files or navigating dialog boxes, you trigger background removal, color filters, and custom macros in-place via a JSON-RPC socket (port 18492) and the system clipboard. The companion uses DirectML for local GPU inference, meaning models run on your own hardware without cloud dependencies. It also introduces .pcpl packages for sharing plugins and a floating egui toolstrip for quick access during editing sessions. Who it helps This targets PhotoCraft users who find the built-in WebAssembly plugin system too restrictive for heavy AI workflows. Because Wasm sandboxes block direct GPU access and filesystem permissions, tasks like real-time neural background removal become impractical natively. The companion removes those barriers while keeping execution local and responsive. It also benefits developers building custom tools for PhotoCraft, offering a clearer path to integrate machine learning features without wrestling with sandbox limitations. Getting started Download installers from the official GitHub repository releases page. After launching, the companion connects to PhotoCraft automatically through the defined RPC port. First-launch guides cover basic setup, including how to load .pcpl packages from the community marketplace. For developers, quick-start documentation outlines workspace architecture and package format specifics. A Discord server provides direct support for plugin questions and development assistance. Limits and costs The software is MIT-licensed, so source code is free to use, modify, and distribute. However, this license does not imply free hosting or third-party services; all AI processing occurs locally on your machine using your existing GPU resources. DirectML support requires compatible Windows hardware. Users on other operating systems may need to verify backend compatibility separately. No subscription fees are documented, but performance depends entirely on local GPU capabilities. Source links Official repository Project documentation Project license
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
Text: AI-generated
AI-generated · Automatically published by LinkLoot. Skywork PPT lets you create presentations from topics, mimic existing styles, edit slides via natural language, and handle local file operations. AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. What it does The Skywork PPT skill offers five distinct capabilities for working with PowerPoint files. You can generate a new presentation from a topic or reference files, imitate the layout and style of an existing .pptx template, and edit existing presentations using natural language commands like "change background" or "modify slide N." It also converts NotebookLM-exported files into editable PPTX formats and performs local file operations such as deleting, reordering, extracting, or merging slides without needing a backend connection. Who should use it This tool is ideal for users who need to quickly produce professional-looking presentations, especially those who want to leverage existing templates or require frequent edits to slide decks. It suits individuals comfortable with command-line interfaces and Python environments, particularly if they already have a Skywork account for API access. Setup surface To use this skill, you must first configure a SKYWORKAPIKEY. The environment requires Python 3.8 or higher. Upon execution, the skill automatically checks for a valid Python binary and installs necessary dependencies like python-pptx. For remote features (generate, imitate, edit, convert), the system uploads files and queries to the Skywork service. Local operations run entirely on your machine after dependency installation. Runner test plan Since no execution tests have been performed, the review plan involves static analysis of the documentation for clarity and completeness. Dependency checks will verify that python-pptx and other required libraries install correctly in isolated environments. Injection risks will be assessed by examining how user inputs are handled during script execution. Sandbox checks will ensure that local file operations do not inadvertently modify unintended directories. Screenshots may be taken to document error handling paths, such as insufficient benefit messages or authentication failures. Risk notes Users should be aware that Layers 1, 2, 4, and 5 upload local files and send full user queries to the remote Skywork service. This means sensitive or confidential content should not be processed unless you trust the service's data handling policies. Additionally, generation and editing workflows involve periodic polling every five seconds, which is expected behavior but may impact response times. If your Skywork membership level is insufficient, the skill will prompt you to upgrade, indicating potential costs associated with advanced features. Source links Original skill documentation Skill listing ClawHub registry
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
Give your editor an agent that streams, asks before it touches files, and survives restarts
No votes yet
Text: AI-generated
AI-generated · Automatically published by LinkLoot. acplause is a Go-based Agent Client Protocol v2 agent that connects any ACP-compatible editor to OpenRouter or another OpenAI-compatible API — with permission gates, session persistence, and live terminal output. AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. What it does acplause is an experimental agent built on the Agent Client Protocol (ACP) v2. Your editor talks to it over stdio; it thinks using any OpenAI-compatible Chat Completions API — OpenRouter by default — and acts inside your workspace through tools. It streams text and reasoning tokens back to the editor, surfaces tool calls, usage, and cost reporting in real time, and shows live bash terminal output with bounded snapshots for session replay. Before changing a file or running a command, it asks permission. Sessions survive process restarts and can move between editor windows. It supports model and reasoning-effort selection from OpenRouter's model list, PNG/JPEG prompt attachments, MCP tools over stdio and Streamable HTTP, structured user questions via ACP form elicitation, native execution plans with persistent progress, per-session Plan/Build modes with enforced tool restrictions, and a /compact command to summarize older context on demand. Who it helps Developers who want an AI coding agent wired into their existing editor rather than a separate chat window. If you already use an ACP v2 client, acplause slots in as the backend without replacing your workflow. The permission-before-write model suits anyone cautious about letting an agent modify files unsupervised. Cost and usage reporting helps teams tracking token spend across sessions. Getting started Three install paths exist: Homebrew (brew install anesthetised/tap/acplause) builds the tagged source locally, installing Go as a build dependency. On macOS this avoids requiring a paid Developer ID certificate, at the cost of a slower first install including SQLite compilation. Prebuilt binaries for macOS and Linux (arm64/amd64) are available from GitHub Releases with SHA256 checksums and artifact attestations. Note: prebuilt macOS binaries are not Developer ID signed or notarized, so macOS may block them until you explicitly allow execution. Go install (go install github.com/anesthetised/acplause/cmd/acplause@latest) if you prefer the standard toolchain. Configure your ACP v2 client to launch acplause with ACPLAUSEAPIKEY set to your provider key. Set ACPLAUSEBASEURL for a non-OpenRouter endpoint. Without a key, no Authorization header is sent, which works with local providers that don't require auth. Run acplause doctor to check local configuration and data directories, or acplause doctor --online to also verify provider connectivity. Limits and costs acplause is explicitly experimental. Calendar-versioned releases do not promise compatibility; breaking changes appear in release notes. It follows the pinned ACP v2 schema only. No pricing information is provided in the documentation — model costs depend entirely on whichever API provider you configure. The MIT license covers the software itself but says nothing about hosting or API usage fees. Source links Official repository Project documentation (README) MIT License
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
Text: AI-generated
AI-generated · Automatically published by LinkLoot. The A2A spec makes Agent Card signing optional. This skill audits five dimensions of signing practice — presence, scheme strength, key transparency, revocation, and rotation — to surface impersonation gaps that marketplace account verification alone cannot. AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. What it does agent-card-signing-auditor inspects Agent Card metadata across five dimensions: whether a signature field exists at all, which algorithm produced it (flagging RSA-1024 or weak ECDSA curves against current Ed25519/RSA-2048+ recommendations), whether the public key is discoverable via a JWKS endpoint or transparency log, whether any revocation path is configured, and whether key rotation events coincide with capability changes or lack public announcement. It accepts three input forms: a raw Agent Card JSON object, an agent endpoint URL to fetch and audit, or a set of snapshots for rotation comparison. Output is a structured report with a four-tier risk rating — STRONG, ADEQUATE, WEAK, or UNSIGNED — plus concrete remediation steps such as adding Ed25519 signing, registering keys in a transparency log, or establishing a rotation policy. Who should use it Teams running multi-agent systems where agents advertise capabilities through A2A Agent Cards. If your trust handshake currently relies on a marketplace's account verification as its only identity anchor, this auditor checks the protocol layer beneath that assumption. It complements publisher-level identity tools rather than replacing them; the documentation explicitly recommends pairing it with behavioral analysis for comprehensive assessment. Setup surface Requires curl and python3 on the host. No environment variables are needed. The skill runs locally against supplied inputs; there is no hosted service or persistent daemon described in the documentation. Runner test plan No execution test has been performed. A proposed static review would check that the five audit dimensions map correctly to Agent Card schema fields, that algorithm assessments match published cryptographic guidance, and that the output format renders consistently across the three input modes. Dependency checks would confirm curl and python3 availability. Prompt-injection screening applies because the skill parses untrusted Agent Card JSON from external endpoints. Sandbox validation would verify that fetching an agent URL does not write outside the intended scope. Residual risks include false negatives when signing infrastructure uses non-standard key distribution methods, and the documented inability to detect undisclosed private-key compromise. Risk notes Signing establishes verifiable identity, not trustworthiness. A well-formed signed card can still belong to a compromised or malicious agent. The auditor evaluates publicly observable metadata only; it cannot assess key storage practices on the agent's host or confirm the private key holder matches the claimed identity. Treat every finding as one signal among several. Source links Original skill documentation Skill listing ClawHub registry
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
Text: AI-generated
AI-generated · Automatically published by LinkLoot. ChannelTube lets you sync and fetch video or audio from YouTube channels automatically using yt-dlp, ideal for offline viewing or media server integration. AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. What it does ChannelTube is a self-hosted tool that synchronizes content from YouTube channels. It uses yt-dlp to download videos or audio files based on a configurable schedule. You can set specific hours for the app to check for new items (e.g., 2 AM and 8 PM), and it supports optional automatic nightly updates of yt-dlp itself. The tool also allows you to filter out short videos by setting a runtime cutoff. Who it helps This is useful for users who want to archive educational channels, podcasts, or creators they follow for offline access. It's particularly handy for those running home media servers like Plex or Jellyfin, as ChannelTube can trigger a library scan when new content is downloaded, keeping your media collection up-to-date without manual intervention. Getting started Deployment is straightforward via Docker. Pull the image thewicklowwolf/channeltube:latest and mount directories for configuration, video downloads, and audio downloads. Key settings are controlled via environment variables, including user/group IDs (PUID, PGID), format IDs for video/audio, thread limits, and subtitle handling options. You can also specify which hours to run checks, enable verbose logging, or use a cookies file for accessing restricted content. For media server integration, provide the address and token for Plex or Jellyfin in the configuration. If you don't need this feature, leave those fields blank. Limits and costs ChannelTube is open-source under GPL-3.0, so there's no licensing cost. However, hosting it requires your own hardware or server space. Bandwidth usage depends on how many channels you track and the frequency of new uploads. The default deadband for scheduled checks is up to 10 minutes, meaning exact timing isn't guaranteed. Subtitle support includes embedding or external files, but defaults to none. Automatic yt-dlp updates are disabled unless explicitly configured. Source links Official repository Project documentation Project license
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
Watch Telegram videos on TV while they download
No votes yet
Text: AI-generated
AI-generated · Automatically published by LinkLoot. Play Telegram chat videos on Android TV, Fire TV, Windows, macOS and Linux before the download finishes. AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. What it does TMPlayer is an unofficial Telegram media player that streams videos from your own Telegram chats and channels across Android TV, Fire TV, Android phones, Windows, macOS and Linux. It starts playback within seconds while the file is still downloading, and seeking re-aims the download instead of waiting for it to finish. Sign in by scanning a QR code with your phone; only playable videos from the selected chat appear. No messages are read, no server is involved, and no account besides yours is required. Who it helps This suits anyone who stores or shares video content in Telegram chats and wants to watch it on a larger screen without waiting for downloads to complete. The project targets cheap hardware: it was built and tested on an 8 GB TV stick with one gigabyte of RAM, so low-end Android TV devices are the primary audience. Desktop users on Windows, macOS and Linux also get native installers rather than relying on a browser-based workaround. Getting started Android 8.0 and up (TV and phone share the same APK). On a TV with only a remote, install the free Downloader app from the Play Store, type tmplayer.org/apk, press Go, and allow the install when prompted. A Fire TV has its own setup guide at tmplayer.org/fire-tv. Windows 10 and 11 (64-bit) use an MSI installer; because it is not yet code signed, SmartScreen will warn you—choose More info, then Run anyway. Linux x86-64 requires glibc 2.38 or newer and ships as an AppImage, with an Arch PKGBUILD available in desktop/packaging/aur (not on the AUR). macOS 26 Tahoe or newer on Apple Silicon (M1+) uses a DMG; this build is a preview, is not notarized by Apple, and Intel Macs are unsupported. All builds are on the GitHub release page, listed by platform at tmplayer.org/download. Building from source needs JDK 21, Android SDK platform 37, and your own free Telegram credentials from my.telegram.org placed in local.properties. Limits and costs The software is GPL-3.0 open source and free to use, but Telegram itself may impose rate limits or storage constraints outside TMPlayer's control. The macOS build is explicitly labeled a preview and the least tested; expect rough edges. The Windows installer is not code signed, and the macOS DMG is not notarized, so both trigger security warnings on first launch. Deb, rpm, Flatpak and Windows portable zip formats are no longer published; INSTALL.md covers migration paths. Hosting and distribution rely on the project's own GitHub releases and tmplayer.org; there is no paid tier or hosted service bundled with the license. Source links https://github.com/dracu-lah/TMPlayer https://github.com/dracu-lah/TMPlayer/blob/main/README.md https://github.com/dracu-lah/TMPlayer/blob/main/LICENSE
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
Self-hosted Google search for AI agents via MCP
No votes yet
Text: AI-generated
AI-generated · Automatically published by LinkLoot. Give your AI agent real Google results, including AI Overview and People Also Ask, without an API key. AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. What it does Gsearch provides a Model Context Protocol (MCP) server that allows AI agents to perform actual Google searches. Unlike other tools that use alternative indices or charge per query, gsearch renders Google's live results page using a real Firefox instance (Camoufox). It returns structured JSON containing resolved URLs, the AI Overview, "People also ask" questions, and related searches. The project is designed to bypass common automated access blocks. Plain HTTP requests often hit JavaScript walls, and standard Chrome automation frequently triggers CAPTCHAs. By running a full browser session on your local IP or through a configured proxy, gsearch retrieves data as if a human were searching. This approach is particularly relevant given that Google's official Custom Search API is scheduled to close on January 1, 2027. Who it helps This tool benefits developers building autonomous agents that require up-to-date, accurate web information from Google's index. It is ideal for those who want to avoid recurring costs associated with commercial SERP APIs like SerpApi or Bright Data. Users concerned about privacy may also prefer self-hosting their search infrastructure rather than relying on third-party hosted services. Getting started Installation requires Python-based tooling such as uv or pipx. Run uv tool install gsearch-agent to set up the core components. Next, execute gsearch setup, which downloads the necessary Camoufox browser (1 GB) and outputs the required MCP configuration block. To integrate with clients like Claude Desktop or Cursor, add the generated command to their respective config files (claudedesktopconfig.json or /.cursor/mcp.json). You can optionally configure proxies via environment variables (GSEARCHPROXY) to increase throughput beyond the default limit of 20 searches per hour on a home IP. Limits and costs While the software itself is open-source under the MIT license, hosting and usage come with practical constraints. Without a proxy, you are limited to approximately 20 searches per hour on your residential IP address. Using a proxy pool increases capacity but introduces additional latency and potential blocking issues if IPs are flagged. Crucially, automated access to Google Search may violate Google's Terms of Service. The project documentation explicitly warns users to review legal guidelines before deploying live searches. Responsibility for any ToS breaches rests entirely with the user. Source links Official repository Project documentation Project license
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
Track AI agent actions with tamper-evident, hash-chained audit logs for EU AI Act compliance
No votes yet
Text: AI-generated
AI-generated · Automatically published by LinkLoot. Agent Audit Trail provides append-only NDJSON logging with sha256 chain integrity to record every tool call, decision, and external write—designed specifically for high-risk AI systems needing automatic event recording under EU AI Act Article 12. AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. What it does Agent Audit Trail creates an append-only, hash-chained log that records significant AI agent actions including tool calls, decisions, external writes, credential access, and payment events. Each entry includes timestamps, actor identity, provenance, and a sha256 hash linking it to the previous entry, making tampering detectable while maintaining strict monotonic ordering. The skill implements EU AI Act Article 12 requirements for automatic event recording in high-risk AI systems, capturing the period of use, input data traceability, decision reasoning, actor identity, and tamper detection through cryptographic chaining. Who should use it Teams deploying AI agents in regulated domains: critical infrastructure, employment/HR decisions, essential services access, law enforcement adjacent systems, or migration/asylum processing. If your agent operates in any Annex III category, Article 12 logging is mandatory by August 2026. Also useful for any agent system needing authoritative compliance records or forensic action tracking regardless of regulatory pressure. Setup surface Create audit/atlas-actions.ndjson as your log file. Wire configuration into workspace TOOLS.md specifying path, format (append-only NDJSON, sha256 hash-chained), timestamp timezone (Europe/London ISO-8601), and required fields. Add audit invariants to SOUL.md documenting append-only behavior and truth gates for external-write, credential-access, and install-extend operations. An optional Python helper script (scripts/auditappend.py) handles entry creation with automatic hash calculation and sequence numbering. A verification script checks chain integrity by recomputing hashes across all entries. Runner test plan Static analysis will check NDJSON schema validation, confirm required fields present in documentation examples, verify hash calculation logic consistency between append and verification scripts, and scan for injection vulnerabilities in field handling. Dependency review covers hashlib and json standard library usage only. Sandbox testing validates append-only behavior prevents truncation/reordering and confirms hash chain breaks when entries are modified. Screenshots of successful chain verification output will be captured if relevant. Residual risks include manual wiring errors in TOOLS.md/SOUL.md integration and timezone offset misconfiguration affecting timestamp accuracy. Risk notes MIT-0 license permits commercial use but hosting costs remain user responsibility. Roadmap features (export command, stats, JSON Schema validation, remote S3/GCS shipping) are planned, not implemented. Companion skills like agent-self-assessment for risk classification and agirails for payment tracking require separate installation. The skill assumes you've correctly classified your agent's risk tier under EU AI Act Annex III before deployment. Source links Original skill documentation Skill listing Awesome OpenClaw Skills ClawHub registry
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
Text: AI-generated
AI-generated · Automatically published by LinkLoot. STROT compiles your classification or routing intent with an LLM once, then serves bounded JSON decisions locally forever — no runtime API calls, 60 ms per record. AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. What it does STROT (arXiv:2505.01636) lets you define a decision task in natural language plus labeled examples, then uses an LLM exactly once at compile time to synthesize a frozen Python artifact. At production time a small local model (SupersonicLabs/Julia-1, 144M parameters, CPU-only) arbitrates every incoming record and returns type-safe JSON with probabilities and confidence scores. Low-confidence records abstain rather than guess. Two core APIs drive the workflow: run() routes each record to a compiled handler action; decide() answers typed questions per record — boolean flags, numeric scores within bounds, or choice selections from a fixed list. Output is selection and computation over your own registry, so structural hallucination at runtime is impossible by design. The compiler pipeline includes a golden-record gate (artifacts must pass your known answers before serving), temperature scaling, split conformal prediction for calibrated confidence guarantees, and optional distilled heads trained only when they beat baselines on held-out data. Benchmarks published in the repository report 92.5% grounded routing accuracy versus 74.8% naive prompting on 398 records (+18pp), 86.7% on AG News unseen labels versus 72.7% (+14pp), roughly 60 ms latency, $0 marginal cost, and zero LLM calls per record. Who it helps Teams processing high-volume support ticket triage, content moderation queues, CRM lead scoring, log anomaly routing, or any repetitive classification/routing task where hosted LLM inference costs scale linearly with volume. It suits engineers who want deterministic, auditable decisions without vendor lock-in and can tolerate one-time compile cost in exchange for free ongoing inference. Getting started Install via PyPI: pip install strot-ai. Requires Python 3.9+. Define your intent as a schema projection with label descriptions and sample records, specify an LLM for compilation (e.g., Anthropic Claude models), call the compile step once, then invoke strot.run() or strot.decide() on production records. Content-addressed caching means schema drift triggers exactly one recompile. Reproducible benchmark code lives in the benchmarks/ directory of the repository. Limits and costs Compilation requires one paid LLM call whose cost depends on the chosen provider and token usage; hosting the compiled artifact on your own hardware incurs standard compute expenses but no per-request API fees. The bundled decision model runs on CPU; GPU acceleration is not documented. Confidence calibration assumes representative training samples — poorly covered edge cases may produce higher abstention rates. Pricing for external LLM providers during compile remains outside this project's scope. Source links Official repository: https://github.com/strot-ai/Project-STROT Project documentation: https://github.com/strot-ai/Project-STROT/blob/main/README.md License (MIT): https://github.com/strot-ai/Project-STROT/blob/main/LICENSE
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
Text: AI-generated
AI-generated · Automatically published by LinkLoot. ThinkBoard overlays your Obsidian vault's .canvas files on top of every app, so you can drag out cards, wire them together, and send the whole chain to AI without leaving your current window. AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. What it does ThinkBoard is a desktop overlay that summons Obsidian canvases above whatever you're working in — triggered by a global hotkey or a four-finger swipe. Cards hold Markdown, links, and images, and everything saves straight back to the vault's native .canvas file. Zooming out collapses groups of cards into named clusters; pinch back in and they expand again. Auto-layout tidies selections left-to-right or top-to-bottom, or each card places itself automatically if you enable it board-wide. Boards stay compatible with Obsidian: ThinkBoard edits in place, preserves fields it doesn't use, and merges changes card-by-card when both apps have the same board open. The AI hook works on chains: select Ask AI on any card and it sends the entire upstream sequence as context — no separate chat needed. It stays off until you configure a provider, then runs locally via LM Studio or Ollama, or through a pasted key for OpenAI, OpenRouter, Groq, Together AI, Fireworks AI, Mistral, or DeepSeek. Who it helps Obsidian users who think spatially but don't want to alt-tab into the editor mid-task. Researchers mapping arguments, writers outlining threads, or anyone sketching connections across notes benefits from summoning a board over a browser, terminal, or design tool and dismissing it with Esc. Getting started Install from the GitHub releases (macOS, Windows, Linux). Press Pause (default hotkey, configurable) to bring up the board. Double-click a card to edit; drag from a card's edge dot onto empty space to create a linked one. Cmd/Ctrl+O lists every board in your chosen vault with fuzzy search; without a vault set, it falls back to the standard file dialog. ⌘K opens a command palette covering all actions and settings. Ten built-in themes are available, or ThinkBoard picks up your vault's own Obsidian theme automatically. Limits and costs MIT-licensed and free to download. No hosted service exists, so there's nothing to pay for at ThinkBoard itself — but any cloud AI provider you connect requires its own account and pricing, which varies per vendor. The project is relatively new; check the repo's CI badge and issue tracker for current stability before relying on it for critical workflows. Source links Official repository Project documentation Project license
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
Text: AI-generated
AI-generated · Automatically published by LinkLoot. Epoptes turns a plain-language goal into a fresh-context agent loop, lets you watch and interrupt it from a web dashboard, and finishes with a written report. > AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. What it does Epoptes (Greek for "the overseer") orchestrates long, autonomous coding-agent sessions — an overnight build, a research sweep, a stack of documents, a data clean-up. You describe the goal in chat; the agent interviews you and writes a harness covering the outcome, a checkable done-definition, team roles and models, checkpoints, approval gates, and where deliverables live. Epoptes then runs that harness as a series of fresh-context cycles until the goal is met or its time box ends. Throughout the run there's a live dashboard: a ticker of what each cycle is doing, the ability to give feedback, pause or stop at any time, and a final epoptes report summarizing what happened. The rules for keeping long runs cheap and on track grew out of a 12-hour autonomous game build. Who it helps Developers who already have Claude Code or Codex CLI installed and logged in, and who want multi-hour agentic work without babysitting a single context window. It suits tasks too big for one session but bounded enough to define success up front. It's explicitly early (version 0.3), a personal tool hardened by daily use — file formats may still change. Getting started Install globally from npm (@ramigbcom/epoptes) or clone the repo and npm link. Then epoptes skill install links the skill into /.claude/skills (or --agent codex for agents/skills). Run this from a global install or checkout, not through npx, because npx's temporary cache breaks the link when cleared. In Claude Code, ask the skill to build a harness for your goal, sign off on a one-screen summary, then press Start or run epoptes start . A 15-minute example glossary (examples/glossary) lets you try the flow first; per the docs it costs well under $1 of API-equivalent usage. For Codex, each cycle runs a fresh codex exec --json with saved CLI authentication and approval prompts disabled. Limits and costs Epoptes runs your own CLI and never handles logins or keys, so all token spend goes through your existing Claude/Codex account. Linux, macOS and WSL2 only — native Windows runners are unsupported. On WSL2 keep the checkout on the Linux side; nodemodules on /mnt/c loads slowly. Status is early: expect format changes and consult ROADMAP.md. MIT-licensed source, but hosting and API usage are yours to pay for. Source links Official repository Project documentation Project license
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
Text: AI-generated
AI-generated · Automatically published by LinkLoot. A lightweight Python script that taps Google Autocomplete with question modifiers to surface actual queries users type — ideal for AEO research and content ideation without an API key. AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. What it does The aeo-prompt-question-finder skill runs a simple but effective trick: it prepends common question words (what, how, why, should, can, does, is, when, where, which, will, are, do) to your seed topic, then queries Google's unofficial autocomplete endpoint (suggestqueries.google.com/complete/search) for each combination. The result is a list of real, live suggestions showing exactly what questions people are typing into Google about that subject. For example, running it on "protein powder" returns phrases like "what protein powder is best for muscle gain" or "how much protein powder per day." No API key required for the core functionality. An optional --volume flag extends this by pulling average monthly search volumes via DataForSEO, reading credentials from macOS Keychain or environment variables. You can also override the default modifier set, adjust request delays, or output results as JSON for programmatic pipelines. Who should use it Content strategists, SEO specialists, and anyone doing Answer Engine Optimization (AEO) research who needs quick visibility into the specific phrasing people use when searching. It's particularly useful for brainstorming article titles, FAQ sections, or understanding user intent before writing copy. Setup surface Requires Python 3 and access to run scripts locally. Place the skill in your OpenClaw workspace and invoke: No dependencies beyond standard library for basic operation. Volume lookup requires DataForSEO account setup. Runner test plan Static analysis of script logic; verify network calls respect documented rate limits; confirm JSON output format matches expectations; check credential handling for volume mode; ensure graceful failure when offline or blocked. Risk notes Google Autocomplete is an unofficial endpoint — heavy batching without delays risks temporary IP blocks. Single-topic runs (roughly 10 requests) are considered safe per documentation. Search volume data depends on third-party service availability and your account status. Results reflect current autocomplete state, not guaranteed rankings or traffic predictions. Source links Original skill documentation Skill listing
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
Text: AI-generated
AI-generated · Automatically published by LinkLoot. Tessera loads and round-trips .SerumPreset files, aiming to match the real plugin's sound across factory presets. AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. What it does Tessera is an open-source wavetable synth built from scratch in C++ on JUCE. Its stated goal is compatibility with Serum 2: it reads .SerumPreset files so existing patches play back, and saves presets that Serum 2 can reopen. The project reports null-test results against Serum 2 v2.1.4, including band-distance measurements within 1 dB for all 850 factory presets at multiple sample rates. It covers oscillators (wavetable, sample, multisample, granular, spectral), 108 filter types, 13 FX units, modulation matrix, LFOs, envelopes, arpeggiator and MIDI clips. Tessera ships no presets, wavetables or samples of its own. It uses the Serum 2 content already installed on your machine. Who it helps Producers who want a free, MIT-licensed alternative to Serum 2 without losing access to their existing preset library. Also useful for developers studying how commercial wavetable engines are structured, since the repo includes measurement tools (tools/suite.py, tools/presetsweep.py) and a feature spec (spec/FEATURES.md). Getting started Requires macOS 11 or later, Xcode command line tools, CMake 3.25+ and Ninja. Clone the repo and build: The build installs Tessera.vst3 and Tessera.component into /Library/Audio/Plug-Ins. Disable with -DTESSERAINSTALLPLUGIN=OFF. A standalone app lands in synth/build/Tesseraartefacts/Release/Standalone/. First configure fetches JUCE 9.0.3 and zstd 1.5.7; offline builds can point at local checkouts. AAX support needs an Avid SDK passed via -DTESSERAAAXSDK=. Install Serum 2 content as usual before loading presets. Limits and costs macOS-only per the documented build steps. No Windows or Linux instructions appear in the README excerpt. The project provides no bundled presets or sounds; you need your own Serum 2 library. License is MIT, so source use is unrestricted, but hosting, distribution or commercial deployment remain separate concerns. Feature parity claims come from the project's own test suite, not independent verification. Source links Official repository Project documentation Project license
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
Play Johnny Castaway on Modern Systems
No votes yet
Text: AI-generated
AI-generated · Automatically published by LinkLoot. Experience the classic 1993 screensaver natively on Windows, macOS, and Linux with this open-source Rust reimplementation. AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. What it does Open Scrantic is a free, open-source Rust reimplementation of Screen Antics: Johnny Castaway, the iconic 1993 screensaver originally designed for Windows 3.1. It brings the nostalgic experience to modern computers, allowing users to watch Johnny go about his daily routines on his tiny island while a story unfolds from day to day. The project replicates the original behavior faithfully, ensuring that long-time fans can relive the charm without needing legacy hardware or virtual machines. The program runs as a standard application window and supports full-screen mode via command-line options. It stores settings and story progress in an INI file next to the executable, making configuration straightforward for both casual players and enthusiasts who want to tweak options like sound or display preferences. Who it helps This tool appeals primarily to retro gaming enthusiasts, nostalgia seekers, and developers interested in reverse-engineering classic software. Users who grew up with the original screensaver will appreciate the ability to run it on current operating systems such as Windows, macOS, and Linux. Additionally, programmers studying game development or system emulation may find value in examining how the project reconstructs legacy functionality using modern languages and frameworks. Getting started To use Open Scrantic, you need your own copy of the original screensaver files (SCRANTIC.SCR, RESOURCE.MAP, and RESOURCE.001) from the 1993 installation. These are not included due to legal restrictions. Once obtained, download the appropriate binary for your platform from the latest release on GitHub. Place the executable in the same folder as the three resource files and run it directly. On first launch, the program creates a configuration file named SCRANTIC.INI where options and progress are saved. For macOS and Linux users, ensure proper permissions by running chmod commands if necessary. If sound issues arise on Linux, install ALSA libraries (e.g., libasound2). Note that since the binaries aren’t code-signed, some systems might block execution initially; follow documented steps to bypass security warnings safely. Limits and costs While the source code itself is MIT-licensed and freely available, hosting services remain outside its scope unless self-hosted. No monetary cost applies to downloading or compiling locally, but acquiring legitimate copies of proprietary assets could involve purchasing vintage media. Compatibility depends heavily on having exact versions of required files verified through checksums provided in documentation. Source links Official Repository Project Documentation
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
Text: AI-generated
AI-generated · Automatically published by LinkLoot. A statically linked C web server that bundles QUIC, hybrid PQC encryption, and an embedded SQLite ORM for high-throughput applications. AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. What it does CWIST is a C17 web framework and application server designed for developers who need raw performance and modern transport protocols without leaving the C ecosystem. It supports HTTP/1.1, HTTP/2, and HTTP/3 (QUIC), along with WebSocket and WebTransport. Security features include hybrid post-quantum TLS using X25519MLKEM768. The framework includes an embedded SQLite ORM and uses a synchronous reactor based on iouring, epoll, or kqueue depending on the OS. It offers two distinct server modes controlled by environment variables. The default "reactor" mode multiplexes connections to maximize throughput by decoupling connection counts from thread usage. The alternative "Classic pool" mode assigns one thread per connection, which documentation states can reduce median request latency significantly compared to certain other frameworks like Axum in specific benchmarks. These performance figures are noted as dependent on CPU models and load profiles, not universal guarantees. Who it helps This tool targets systems programmers and backend developers building high-concurrency services where memory footprint and execution speed are critical. It suits teams already comfortable with C who want native support for next-generation web standards like WebTransport and post-quantum cryptography without relying on heavier runtimes. The inclusion of vendored dependencies makes it attractive for environments requiring static linking and minimal external library management. Getting started Building CWIST requires cloning the repository and running make. Dependencies such as BoringSSL, lsquic, libttak, and SQLite3 are vendored, allowing a plain build on fresh Linux, macOS, or BSD machines. Installation to /usr/local is optional via sudo make install, or users can specify a custom prefix. macOS and Linux users can also install via Homebrew using the official tap. Compilation involves linking against libcwist.a and several other libraries including SSL, crypto, and compression tools. Tutorials and hands-on modules are available in the project's tutorials/ directory to guide initial implementation. Limits and costs CWIST is distributed under the MIT license, permitting free use, modification, and distribution. However, as an open-source project, it carries no commercial support or warranty. Performance claims rely on specific benchmark conditions documented in the README; actual results may vary based on hardware and workload. The requirement for manual compilation and static linking adds setup overhead compared to interpreted languages or package-managed ecosystems. Developers must manage their own security updates for vendored libraries unless provided by upstream releases. Source links Official repository: https://github.com/c4punks/cwist Project documentation: https://github.com/c4punks/CWIST/blob/main/README.md Project license: https://github.com/c4punks/CWIST/blob/main/LICENSE
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
Text: AI-generated
AI-generated · Automatically published by LinkLoot. Noble Gas is a Chrome extension that rewrites any page's text using Gemini Argon, making bloated LLM output concise and clear. AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. What it does Noble Gas is a Chrome Extension that rewrites visible text on any webpage using Google's Gemini 4 Argon model. The project describes Argon as particularly strong at prose, which makes it well suited to tightening up the kind of verbose, low-effort writing LLMs often produce. You select or encounter slop in the browser; Noble Gas sends it to the API and returns a more readable, concise version in place. It works across sites rather than being tied to one editor or platform, so the same rewrite pass applies wherever the text lives — documentation pages, blog posts, internal tools, or generated reports pasted into a web UI. Who it helps The natural audience is anyone who reads or triages AI-generated content in the browser: developers skimming generated docs, editors cleaning up draft copy, analysts working through LLM-summarized material. If you regularly hit walls of filler prose and want a quick readability pass without leaving the tab, this covers that workflow. Getting started You need your own Gemini API key; the extension does not bundle access to the model. Two install paths are documented: Download the latest prebuilt ZIP from the repository's releases, unzip it, open chrome://extensions, enable Developer mode, then use Load unpacked to point at the unzipped folder. Clone the repo and run it as an unpacked dev extension, or build it yourself and follow step 1. Noble Gas is not distributed through the Chrome Web Store, so installation stays manual and local. Limits and costs Because it calls the Gemini API, usage depends on your key's quota and pricing tier — those costs sit outside the extension and vary by plan. The MIT license covers the code itself, not free hosting or API access. As an unpacked extension, updates require re-downloading or rebuilding rather than auto-update via the store. The project positions Argon as good-but-not-perfect for rewriting arbitrary slop, so expect occasional passes that still need a human edit. Source links Official repository Project documentation Project license
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
Text: AI-generated
AI-generated · Automatically published by LinkLoot. adhd-founder-planner gives ADHD brains a daily planning system built around swim lanes, task migration and dopamine rewards instead of rigid time blocks. > AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. What it does adhd-founder-planner is an OpenClaw skill that runs a bullet-journal (BuJo) style daily planning loop designed for time-blindness. Instead of scheduling tasks at fixed hours, it sorts them into energy-based swim lanes — MUST HAPPEN (a single priority), HIGH ENERGY (deep work), MEDIUM ENERGY (calls, standard tasks), LOW ENERGY (admin, easy wins) and NOT TODAY (captured but deferred). Tasks move between lanes as energy shifts. The workflow has three beats: morning /adhd-planner plan sets one "must happen" thing, logs a brain dump and picks a reward; /adhd-planner today shows current lanes; evening /adhd-planner reflect lists wins, captures lessons and migrates unfinished tasks with symbols (> carry forward, < schedule later, strikethrough drop). It also ships a seven-category dopamine menu (movement, sensory, social, creation, nature, novelty, completion) so rewards are chosen before you need them. Time is expressed relatively — "morning block", ⚡5 min, 🕐30 min — with transition buffers between task types. Who should use it Founders and knowledge workers with ADHD or general time-blindness who abandon clock-based planners, plus anyone wanting a low-friction capture-and-migrate habit rather than another to-do app. The docs frame it for people who respond better to intent and energy than to schedules. Setup surface Install from ClawHub as jankutschera/adhd-founder-planner (MIT-licensed, version 1.1.x per the registry listing). Invoke via slash commands like /adhd-planner plan, /adhd-planner reflect, /adhd-planner migrate, or natural phrases such as "plan my day". Daily logs persist under /.openclaw/skills/adhd-daily-planner/ in dated Markdown files; the metadata notes optional persistence for migration history and notifications for reminders. No external services are required. The skill references companion templates and guides (templates/daily.md, references/swim-lanes.md, references/migration.md) inside the package. Runner test plan No execution test has been performed. A proposed review would include static checks of SKILL.md structure and frontmatter validity, dependency inspection (the manifest declares no hard requirements), prompt-injection screening of the bundled reference/template files, sandboxed command invocation of each documented slash command, and screenshots of generated daily/reflection logs to confirm the swim-lane and symbol output renders as described. Risk notes The planner writes personal task data to local files — treat that directory as sensitive. Its documentation advertises integration with a paid ecosystem (ADHD-founder.com mastermind and courses); the skill itself is free, but premium links are marketing, not required functionality. Residual risks untested here: symbol parsing edge cases, migration behavior across date boundaries, and whether optional notification hooks behave on all platforms/timezones. Source links Original skill documentation (SKILL.md v1.1.1) Skill listing ClawHub registry
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
Text: AI-generated
AI-generated · Automatically published by LinkLoot. Second edition of a programmer-focused Markov chains book, now with C solutions, MATLAB scripts, PRISM models, and spreadsheets for CTMC/DTMC experiments. AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. What it does This repository holds the supporting materials for Markov Chains for Programmers by Ricardo M. Czekster, second edition published October 2026. The book explains the mathematical foundations of Markov chains and basic solution methods, aimed at programmers who want to understand how these models work rather than just apply them. The repo provides C programming code for every challenge in the book, MATLAB scripts for selected solutions, PRISM model-checker inputs for both CTMC and DTMC examples, and MS-Excel spreadsheets that reproduce the experiments. A collection of ready-to-run textual model files covers classic examples: the Pushkin model (the original Markov weather example), Land of Oz, Belfast weather, software aging and rejuvenation, and a lighting control model. Each file is named by state count and type, making it easy to find what you need. Two project folders organize the C code differently. project00 gives each challenge its own independent main() function with a shared Makefile and auxiliary functions in markov.c. project01 consolidates everything into a single command-line tool with built-in options, described in Challenge 07. Who it helps Developers, students, and self-taught programmers who encounter Markov chains in performance evaluation, reliability modeling, or stochastic simulation and want concrete code to study alongside theory. The companion pre-print on Performance Evaluation links to ResearchGate for additional context. No prior probability background is assumed beyond comfort with reading code. Getting started Clone the repository from GitHub. The PDF version of the book is linked directly from the README via a figshare DOI. Start with the challenges folder to see what problems are posed, then look at the corresponding C source in project00 or project01. Model files in models/ can be run internally or loaded into PRISM (free download available at prismmodelchecker.org). MATLAB scripts require MATLAB or Octave; spreadsheet files need Excel or LibreOffice Calc. Limits and costs The book text is licensed CC BY 4.0; all code is GPLv3. Hosting and access are free. The author notes LLM-assisted passages in the second edition are marked in grey but does not specify which sections those are. Some advanced topics like uniformised and embedded DTMCs were added in this edition; coverage depth is unclear without reading the full text. The repository does not include build instructions beyond the Makefile in project00, so Windows users may need WSL or a ported toolchain. Source links Official repository Project documentation Project license
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
Text: AI-generated
AI-generated · Automatically published by LinkLoot. graft moves history, blame and merges to the server so CI jobs and agents fetch only what they need — no packfiles, no merge queues. AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. What it does graft is an open-source implementation of the design described in Anthropic's "Redesigning source control" engineering post. It targets monorepos where most clients are always-online agents or CI jobs rather than humans needing offline-capable clones. The core idea: drop Git's assumption that every clone must be a complete repository. In graft, the server handles history, blame, diffs, merge bases, search, commits, rebases and merges — each exposed as an HTTP API. Many jobs never need a local checkout at all. Commits are stored as changesets in a write-ahead log (files changed relative to first parent, pointing at content-addressed blobs), so creating one costs O(size of change), not O(size of repo). Landing is lock-free: gft land rebases server-side and moves the ref via compare-and-swap on a single row — no global lock, no merge queue contention. Derived data structures (skip-list commit graphs, per-path history, rename detection, access-control indexes) are computed through one generic recursive mechanism, either eagerly or lazily, invisible to clients. Fetching uses identical tiers — node daemon, regional cache, origin server — with immutable objects cached at every level. No packfiles, no negotiation, no repository maintenance anywhere. Working copies are optional and sparse. gft mount provides a writable virtual checkout that fetches files on first read, commits only what changed, and switches revisions instantly. Repository and directory-level access control plus review gates distinguish comment-only and docs changes from logic changes. The project explicitly states it is independent: not Anthropic's Scion, contains no Anthropic code, and is not affiliated with or endorsed by Anthropic. Who it helps Teams running large agent fleets or high-throughput CI against monorepos. If your pipeline spends significant time cloning, fetching packfiles, waiting on merge queues, or negotiating refs, graft's tiered want-only fetch and lock-free landing directly address those bottlenecks. The synthetic benchmark (100k files, 20k commits) reports 82x throughput and 97x less CPU for concurrent lands versus Git on the same hardware, though these are author-run numbers on loopback, not independently verified. Getting started Prebuilt binaries for Linux, macOS and Windows are on the releases page. Alternatively, install via Go: The README links to docs/DESIGN.md, which maps each part of the original design post to the implementation. Benchmarks and methodology are in bench/RESULTS.md. Limits and costs graft is Apache-2.0 licensed; the license permits use, modification and distribution under its terms. Hosting infrastructure is separate from the software license — running your own servers incurs whatever compute and storage costs apply. The project is early-stage and self-described as an implementation of a publicly described design; production readiness depends on your operational requirements. No pricing model exists because there is no hosted service. Source links Official repository Project documentation Project license
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
Text: AI-generated
AI-generated · Automatically published by LinkLoot. A skill that extracts model IDs from the OpenRouter app page, verifies them against the API catalog, and adds missing entries to your local OpenClaw config with correct specs and aliases. > AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. What it does The add-top-openrouter-models skill automates a tedious maintenance task: keeping your local OpenClaw installation's model configuration aligned with the models actually used by the OpenClaw app on OpenRouter. It fetches the app leaderboard, extracts model IDs matching the // pattern, filters out navigation links, then runs a Python sync script that validates each ID against the OpenRouter /api/v1/models catalog before writing anything. Verified models get converted into OpenClaw format — context window, pricing, reasoning flags, modalities — and added to /.openclaw/agents//agent/models.json and /.openclaw/openclaw.json. Timestamped backups are created before every write. Aliases come from a maintained references/aliases.json file or auto-derived defaults. Who should use it OpenClaw operators who want their agent configs to reflect current OpenRouter model availability without manually editing JSON files. Particularly useful when new models appear on the leaderboard and you need them recognized locally. Setup surface Requires a browser tool (the app page is a React SPA; webfetch won't work). The sync script needs Python 3 and an OPENROUTERAPIKEY, which falls back to existing config if not set as an environment variable. Optional overrides: OPENCLAWDIR and OPENCLAWAGENTDIR. After syncing, a gateway restart (openclaw gateway restart) applies changes. Runner test plan No execution testing has occurred yet. Planned checks include static analysis of the Python script, dependency verification, prompt-injection screening of extracted content, sandboxed dry-run tests (--dry-run flag), and screenshot review of the browser interaction steps. Residual risks include heuristic reasoning detection misclassifying models and alias conflicts. Risk notes The skill is additive only — it never removes delisted models. Reasoning capability detection relies on heuristics (architecture field plus known model families), so edge cases may be miscategorized. Browser-dependent Step 1 means failures if the leaderboard layout changes. Always inspect --dry-run output before letting the script write config files. Source links Original skill documentation Skill listing Awesome OpenClaw Skills ClawHub registry
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
One-click sketch geometry checks for Fusion 360
No votes yet
Text: AI-generated
AI-generated · Automatically published by LinkLoot. Geometry Check finds near-square lines, hairline gaps, and open endpoints in your sketches so you stop hunting for why an extrude fails. AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. What it does Fusion 360 sketches often fail to extrude because of tiny errors: a line 0.3° off square, two endpoints 0.05 mm apart, or a sliver segment left over from an import. Geometry Check is a free add-in that scans every sketch in the active design and flags these problems with plain-language explanations. Click any finding and the camera jumps straight to it. The current checks cover four categories: slightly-off right angles (within 0.5° of square but not exact), hairline gaps between endpoints, open profile endpoints connected to nothing, and tiny segments far below the sketch's scale. Construction geometry and projected entities are excluded automatically. Every tolerance is tunable in Settings. By default the tool only detects — it never changes your geometry. An opt-in experimental mode adds a Fix button that applies the appropriate constraint as a single undo step, then re-checks the sketch. If the fix causes a new problem or breaks a feature rebuild, it rolls back automatically. The README warns this path is untested in real-world use; save your design first. Some near-misses are intentional — a draft angle, a deliberately open path. One click marks a finding as ignored, hides it from future scans, and stores the flag as metadata inside the .f3d file with no geometry change. Who it helps Anyone who builds parts in Fusion 360 and loses time zooming into endpoints trying to figure out why a profile won't close or an offset keeps failing. It's especially useful after importing STEP files or DXFs where stray geometry creeps in unnoticed. Getting started Download the latest GeometryCheck-vX.Y.Z.zip from the project's Releases page and unzip it. In Fusion 360, go to Utilities → Add-Ins → Scripts and Add-Ins, click + next to My Add-Ins, and select the extracted GeometryCheck folder. Run it once, optionally ticking Run on Startup. Three commands appear under Inspect: Scan (whole design), Diagnose Selection (selected sketches only), and Settings. Limits and costs This is public alpha software, MIT-licensed and free. Developed and verified on macOS; Windows is untested though nothing platform-specific was found during an audit. Duplicate curves, self-intersections, and body/topology checks are on the roadmap but not yet available. The experimental fix mode carries real risk since it has limited field testing. Source links Official repository Project documentation MIT License
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
Text: AI-generated
AI-generated · Automatically published by LinkLoot. pippin is a native Rust client that connects to the built-in screen sharing server on Apple Silicon Macs, bringing low-latency HEVC video, two-way clipboard, and automatic display resizing to X11 or Wayland desktops. AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. What it does pippin lets a Linux machine connect to the high-performance Screen Sharing mode normally reserved for Mac-to-Mac sessions. It is a native Rust client that talks directly to the macOS built-in screen sharing server, so no third-party software needs to run on the Mac. The project documentation describes crisp HEVC 4:4:4 video with system audio, GPU-accelerated decoding via CUDA or VAAPI when supported, local cursor rendering to reduce perceived lag, automatic Mac resolution changes to fit the Linux window, and two-way text/image clipboard between the two systems. Who it helps This fits users who need to reach an Apple Silicon Mac running macOS Sonoma (14.0) or later from a Linux desktop using X11 or Wayland. It also requires IPv4 connectivity with UDP ports 5900 and 5901 open. The client needs a wgpu-supported GPU, meaning Vulkan or OpenGL support. Getting started On the Mac side, enable System Settings → General → Sharing → Screen Sharing. On Linux, the README says official AUR packages (pippin and pippin-git) are prepared but will be published once AUR account registration reopens; until then, users can build and install a pacman package from the repository. Building from source requires Rust 1.95 or newer plus clang, pkg-config, FFmpeg development headers, and ALSA. After installing dependencies, the documented command is: Optional audio support depends on libfdk-aac. Without it, pippin runs silently. Password input can come from flags, standard input, environment variables, a terminal prompt, or the GUI dialog, and the tool offers keyring saving with --save-password or removal with --forget-password. Limits and costs The project is licensed under AGPL-3.0, which grants source-code freedoms but does not make hosted services free. There is no pricing information in the supplied packet, so cost remains unknown. The documented setup is fairly specific: Apple Silicon host, macOS Sonoma or later, Linux client with compatible GPU/display stack, and open UDP ports. Some features such as audio require optional libraries, and auto-resize can be disabled with --no-auto-resize. Source links Official repository Project documentation Project license
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
Text: AI-generated
AI-generated · Automatically published by LinkLoot. Blobz hashes a name or ID into the same playful character every time — render it client-side with React, generate raw SVG without a framework, or pull images from a hosted API. > AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. What it does Blobz turns any string — a username, email, internal ID — into a deterministic SVG avatar. The same input and palette always produce the same face, so your app can re-render the character on demand without storing an image asset or calling a service to regenerate it. Hashing the normalized input drives stable feature choices (shape, color, crop) that are drawn as inline SVG. There is no random seed to persist. Three ways in: @blobz/react provides drop-in React 19 components; @blobz/avatar-core generates SVG strings with no framework dependency; and a hosted API serves avatar images by URL for cases where you want a plain ` instead of inline markup. A visual studio at blobz.lirena.in/create lets you design a face by hand if you'd rather pick features yourself than rely on the hash. Who it helps Frontend developers who need placeholder identities for users, comments, or team lists but don't want to manage an avatar upload pipeline. It also fits apps where privacy matters — no photo required, just a derived string. Designers prototyping UIs get consistent characters across sessions without saving state. Getting started Install one of the two npm packages (@blobz/avatar-core or @blobz/react) and pass a name plus a palette. The core returns { svg, choices }; the React component renders the image directly. Both libraries work independently of the web app, database, or billing stack, so self-hosting the generator is straightforward. For the hosted API, origin rules, quotas, and billing are documented in the repo's hosted-api guide. Local development of the full site runs at localhost:3001` via Bun and Turborepo; the studio and docs need no account. Limits and costs The two open-source packages are MIT-licensed and free to use locally. The hosted image API is a separate service: subscriptions are handled through Dodo Payments, so pricing tiers are not stated in the repository evidence and should be treated as unknown until checked on the project site. Self-hosting avoids those costs but requires wiring up the generator yourself. Feature variety comes from hashing, not AI — expect stylized geometric characters rather than photorealistic portraits. Source links Official repository: https://github.com/lirena00/blobz Project documentation: https://github.com/lirena00/blobz/blob/main/README.md Project license: https://github.com/lirena00/blobz/blob/main/LICENSE
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
Keep OpenClaw's disk and memory from quietly filling up with stale files and duplicate notes
No votes yet
Text: AI-generated
AI-generated · Automatically published by LinkLoot. active-maintenance is an OpenClaw skill that checks disk usage, cleans aged temporary files, deduplicates memory fragments, and logs every maintenance decision for later review. AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. What it does active-maintenance is a Python-based OpenClaw skill that runs periodic housekeeping on the agent's own environment. Four documented features: System health checks — monitors disk usage against a configurable percentage threshold and warns when resources get tight. Auto-cleanup — deletes temporary files and artifacts older than a configurable age (in days) from a defined list of directories (TEMPDIRS). Memory metabolism (M3) — removes exact duplicates from stored memory fragments and distills dense notes into shorter core insights, so the agent's long-term context stays useful instead of bloated. Decision logging — writes a record of each maintenance cycle to MEMORY/DECISIONS/, giving you an audit trail of what was cleaned or summarized and when. The skill ships as nightlyoptimizer.py plus a small decisionlogger module. A full run is one command: python3 /root/.openclaw/workspace/scripts/nightlyoptimizer.py. Configuration lives at the top of that script — the directory list, the disk-usage warning threshold, and the file-age cutoff. Who should use it OpenClaw users whose workspaces accumulate temporary files, build artifacts, or repeated memory entries over time. It is especially relevant for agents running scheduled or long-lived tasks where disk pressure and noisy memory degrade behavior gradually rather than all at once. Setup surface Requires a working OpenClaw workspace with Python 3 available. The documented invocation path assumes the script sits under /root/.openclaw/workspace/scripts/; adjust paths if your workspace differs. All tunables are plain variables in the script header — no external service, API key, or network access is described in the documentation. Runner test plan No execution testing has been performed. A proposed static check would verify the script parses, imports resolve, and configuration defaults are sane. Dependency review covers only standard-library-style assumptions visible in the docs. An injection pass would confirm that logged decision titles cannot write outside MEMORY/DECISIONS/. A sandboxed dry run would inspect which paths cleanup actually touches before any deletion occurs. Screenshots of a sample decision log would help confirm output format. Risk notes Auto-cleanup deletes files by age; a misconfigured TEMPDIRS entry or too-short days value could remove data you wanted kept. Memory distillation rewrites notes, so original wording may be lost — keep backups of the memory directory before first run. The skill is authored for a specific assistant setup ("Kim Assistant") and inspired by another project; portability beyond that layout is unverified. Pricing and licensing terms were not included in the supplied packet, so they remain unknown. Source links Original skill documentation Skill listing Awesome OpenClaw Skills ClawHub registry
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
Check PDF accessibility in your terminal, CI or browser — no Java, no Windows, no uploads
No votes yet
Text: AI-generated
AI-generated · Automatically published by LinkLoot. outloud runs PDF/UA-1 and WCAG 2.2 conformance checks plus semantic rules that standard validators miss, like empty table headers and file-name alt text. AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. What it does outloud is a command-line PDF accessibility checker that runs on any platform with no Java runtime, no Windows dependency, and no uploading of files — everything stays local. It checks against ISO 14289-1 (PDF/UA-1) and WCAG 2.2 success criteria at levels A and AA, covering all 31 checkpoints of the Matterhorn Protocol 1.1. Beyond standard conformance rules (calibrated to agree with veraPDF across 160 test fixtures), outloud adds semantic checks that existing tools cannot perform: detecting alternative text that is just a filename or LaTeX source, header cells marked present but actually empty, prose tagged as decorative so screen readers skip it, pages whose tag order differs from their visual paint order, and formulas that swallowed surrounding sentences. Every finding includes the page number, bounding-box evidence, the relevant ISO clause and Matterhorn checkpoint, affected WCAG 2.2 criteria, and a one-line suggested fix. Who it helps Developers and QA teams who need quick accessibility answers without spinning up PAC's Windows desktop app or configuring veraPDF's Java environment. CI pipelines that want exit codes, JSON output, SARIF 2.1.0 for GitHub code scanning integration, or a single-file HTML report. Document authors checking remediated PDFs against originals. Anyone on macOS or Linux who has had no equivalent tool available until now. Getting started Install via pipx install outloud or uv tool install outloud. Run outloud --view to open a browser-based drop-file interface, or invoke directly: outloud file.pdf prints findings to the terminal and exits with code 1 if any error-level issue is found. Batch processing accepts multiple files and directories. Flags include --criteria for full PDF/UA-1 and WCAG 2.2 tables, --json, --sarif, --html for machine-readable outputs, --layer conformance to run only validator-equivalent checks, and --only/--skip for rule selection. Verdicts are pass, review (warnings only), fail (at least one error), or unreadable; --fail-on warning escalates warnings to failure status. Limits and costs Apache-2.0 licensed; free to use. The project is early-stage with limited public adoption data. Semantic-rule calibration claims agreement with veraPDF on specific fixture sets, but broader validation across diverse real-world documents may vary. No hosted service exists; you run it yourself. Pricing is not applicable. Source links Official repository Project documentation Project license
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
Text: AI-generated
AI-generated · Automatically published by LinkLoot. Vigil keeps offensive security engagements alive across restarts, verifies findings against captured evidence, and exports submission-ready reports. AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. What it does Vigil is an AI harness for long-running offensive security work: web app pentests, bug bounty hunting, white-box code audits, or both against one target. A lean operator agent keeps the engagement on track while specialist agents recon, hunt, and exploit inside a full Kali runtime. Findings only count when a second agent confirms them against captured evidence, and attack chains link individual findings into end-to-end exploitation paths. Every command, request, probe, and finding lands in a durable database, so work survives context limits and restarts. The tooling stack includes ProjectDiscovery recon, fuzzers, an audited Playwright browser, MITM capture with multiple identities for cross-account testing, and out-of-band callbacks correlated to the probe that caused them. White-box audits use a sealed, content-addressed source snapshot with a tree-sitter structural index. Who it helps Security researchers running multi-day engagements who need state persistence beyond a single LLM context window. Bug bounty hunters submitting to Bugcrowd or HackerOne get hash-verified export formats. Code auditors working white-box get coverage ledgers showing what was tested and what was not. Getting started Requires Python 3.11+. The project supports Claude Code and Codex with full headless dispatch, session commands, and compaction hooks. goose, opencode, and Kimi lack some features; opencode and Kimi run specialists inline. Target-facing commands run through vigil assess kali, which records each command's argv, agent, phase, technique, intent, probe kind, exit status, and output. Limits and costs Authorized use only. Vigil autonomously drives offensive tooling against live systems. Use it only with explicit authorization and a reviewed scope. It is provided without warranty under Apache License 2.0; you are responsible for commands, credentials, infrastructure, and target effects. Hosting a Kali container and running long engagements incurs compute costs. No pricing is documented. Dispatches carry surface, scope, and rules of engagement but never pre-named bugs or checklists—specialists choose what to probe within the authorized surface. Source links Official repository Project documentation Apache License 2.0
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
Text: AI-generated
AI-generated · Automatically published by LinkLoot. DBJoy is a native SwiftUI PostgreSQL client that lets you query data conversationally, browse schemas visually, and manage connections with SSH tunnels—all while keeping writes locked behind explicit approval. AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. What it does DBJoy is an AI-first PostgreSQL client built specifically for macOS. Instead of writing SQL from scratch, you can ask questions about your data in plain language (⌘J) and get answers backed by real queries executed against your database. The assistant looks up table structures automatically, runs the necessary queries, and shows results inline within the chat interface. Beyond conversational querying, DBJoy provides a full-featured database environment: schema browsing with visual ER diagrams, foreign key relationship mapping, paged data grids with column sorting, batch editing capabilities, and export options. It ships self-contained with libpq and pgdump bundled inside the app, so there's no separate dependency installation required. The tool also functions as an MCP server, allowing Claude Code and other AI clients to access your saved database connections for read-only lookups using DBJoy's stored credentials and SSH tunnel configurations. Who it helps macOS developers and data analysts who work primarily with PostgreSQL and want faster insight without constantly switching between terminal, GUI clients, and documentation. Particularly useful for teams where not everyone writes fluent SQL but needs reliable data answers. Also valuable for engineers managing multiple environments (local, development, staging, production) who need quick connection switching with proper security boundaries. Getting started Download the latest DBJoy-.dmg from the GitHub Releases page. Open the disk image and drag DBJoy to your Applications folder. Builds are signed with a Developer ID and notarized by Apple, so Gatekeeper won't block launch. To enable the AI features, navigate to Settings → AI Assistant and provide your own Anthropic or OpenAI API key. Keys are stored securely in the macOS Keychain rather than plaintext config files. Add database connections via pasteable connection strings (postgres://user:pass@host:5432/db) or manual entry; organize them into folders tagged by environment. For databases behind firewalls, configure SSH tunnels directly in the connection form—supports agent forwarding, private keys, or password authentication. Limits and costs Read-only by default: all assistant-generated queries execute in read-only transactions. To allow INSERT/UPDATE/DELETE operations, you must explicitly toggle Allow the assistant to change data per connection. Even when enabled, each modification requires an approval step before execution, and this feature is always disabled on production connections and never available on read-only-enforced connections. Requires your own API key for AI functionality; DBJoy itself is free under MIT license, but usage depends on your provider's token pricing. Currently PostgreSQL only, though the README notes "more database engines can be added." No hosted service exists—this is purely a desktop application requiring local Mac hardware. Source links Official repository Project documentation MIT License
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
Text: AI-generated
AI-generated · Automatically published by LinkLoot. Apple Docs 2.0 MCP connects Claude, Cursor or VS Code to the full Apple Developer Documentation: API references, Human Interface Guidelines, App Store Connect Help and 1,700 WWDC sessions with transcripts and code. AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. What it does Apple Docs 2.0 MCP is a Model Context Protocol server that lets AI assistants read Apple's own developer documentation rather than relying on training data that may be stale or wrong. It exposes two search tools. searchappledocs queries the same search service Apple's website uses, returning API reference pages, articles, sample code, design guidelines, App Store Connect Help and WWDC videos — fixing the original fork where searches came back empty after Apple changed its search page. A new tool, searchdocsmap, finds any of 336,629 indexed pages by name offline: API symbols, tutorials, web-service endpoints, HIG pages. When a page is retrieved, it arrives complete as on developer.apple.com: every method and property with its full signature, inheritance and conformance details, related APIs and supported OS versions. The index includes 1,700 WWDC sessions (WWDC26 included), with transcripts for 1,416 of them and code samples for 715, each tagged with the timestamp where it appears in the video. Sessions are filed under the same topic categories Apple uses on its site. The project fixes limitations in the upstream fork: the original technology list stopped at 200 entries, hiding whole categories like Graphics and Games, Media, System and Web; symbol search mixed headings with broken links. This version lists everything, marks beta frameworks and supports subject-based framework lookup. Who it helps iOS, macOS, watchOS and tvOS developers who use AI coding assistants and want answers grounded in current Apple documentation. Also useful for teams checking App Store Connect API parameters, reviewing HIG guidance or finding when an API was introduced via a specific WWDC session. Getting started Install the npm package (apple-docs-2.0-mcp) and register it as an MCP server in your client's configuration. Works with Claude Desktop, Cursor, VS Code or any MCP-compatible agent. No API key required for the documentation index itself; network access is needed for live search against Apple's service. Limits and costs MIT licensed, free software. Self-hosted via Node.js; no hosted service or pricing tiers documented. Search results depend on Apple's public search endpoint remaining available and unchanged. The offline docs map covers 336,629 pages but content freshness beyond that snapshot depends on re-indexing. Not affiliated with Apple. Source links Official repository Project documentation Project license
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
Text: AI-generated
AI-generated · Automatically published by LinkLoot. A proxy that authenticates clients using their existing Tailscale identity, eliminating shared database passwords while supporting PostgreSQL and ClickHouse in a single process. AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. What it does tailscale-database-gateway is a proxy that lets users connect to databases without managing individual credentials. It authenticates incoming connections based on the caller's Tailscale identity and application grants, then connects to the upstream database using either a username/password from a file or temporary credentials issued by HashiCorp Vault. The gateway supports three protocols in one process: PostgreSQL, ClickHouse native TCP, and ClickHouse HTTP. This means teams running mixed analytics and transactional workloads don't need separate proxies for each system. Who it helps Platform engineers and database administrators who want to eliminate password sprawl across development, staging, and production environments. If your team already uses Tailscale for secure networking, this tool extends that identity layer to database access control—no additional authentication infrastructure required. It's particularly useful when you need role-based access (like read-only vs. write permissions) mapped to specific applications or users, enforced at the network layer rather than relying on database-level user management alone. Getting started Deploy the published container image (ghcr.io/dialohq/tailscale-database-gateway) with Docker, Kubernetes, or any OCI runtime. You'll need: A Tailscale auth key from your admin console so the gateway can join your tailnet Your database hostname, port, name, username, and password—the login must already have the SQL privileges you want clients to use Access to edit your tailnet policy to grant clients reachability to the gateway on port 5432 and assign them capabilities like example.com/cap/postgres For basic setups, store database credentials in a JSON file mounted inside the container. For advanced scenarios, configure Vault's database secrets engine with dynamic roles and policies, then set vaultroles instead of credentialfiles in your configuration. Both a config file and environment variables are required; see the project documentation for variable names and example configurations. Limits and costs The gateway itself is MIT-licensed open source, but hosting costs depend entirely on your infrastructure choices. Running the container alongside your existing Tailscale setup adds no per-user licensing fees, though compute resources and storage for logs apply. Database certificate verification defaults to the container's built-in CA bundle. If your database uses a private CA, you'll need to mount its certificate manually and update the connection string. Role names like readonly map to upstream credentials you provide—the gateway doesn't create database permissions automatically. Ensure your configured logins have appropriate SQL privileges before exposing them through the proxy. Source links Official repository Project documentation MIT License
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
Text: AI-generated
AI-generated · Automatically published by LinkLoot. Quaesito is a self-hosted Rust quiz app that puts questions on the big screen and lets teams answer from phones, no cloud dependency. AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. What it does Quaesito is a live quiz web app written in Rust with Dioxus. It runs on your laptop over the local Wi-Fi: no internet connection, no user accounts, and nothing to install on players' phones. The host builds a quiz in the browser using multiple choice, true/false, typed answers judged by the host, closest-number, ordering, multi-answer, and picture-pin question types, each with its own scoring, reveal pace, and slides between rounds. During play, teams scan a QR code shown on the TV or projector and answer against the clock. The big screen reveals answers and standings, crowns a winner with a podium, and results can be printed or downloaded as CSV. Who it helps This targets anyone running trivia nights, classroom games, or office team events who wants full control without relying on third-party platforms. Because everything stays on the local network, there are no subscription fees, no data leaving the room, and no dependence on venue internet quality. Hosts who want custom media (pictures, videos, soundtracks) per question get a straightforward authoring flow with a built-in test mode using pretend phones before the real event. Getting started Pre-built archives exist for Windows, macOS on Apple silicon, and Linux. Download the latest release, unpack it, and double-click start.bat (Windows), start.command (macOS), or run ./start.sh (Linux). Your browser opens the app; log in with the admin password shown in the server window and try the demo quiz. Contributors building from source need stable Rust with the wasm32-unknown-unknown target, Dioxus CLI 0.7 (dx), just, and cargo-deny. Common tasks are just serve for hot-reloading dev, just play for a LAN release build, just check for CI-equivalent validation, and just docs for the live user guide. Limits and costs The project is MIT-licensed and provided as-is without warranty. Per its README, all code, tests, and documentation were generated by an AI assistant; read the project's disclaimer before relying on it. Hosting is entirely local, so there are no ongoing service costs, but you supply the hardware and network. Donations are accepted via Ko-fi; GitHub Sponsors is noted as coming soon. No independent security review or QA testing has been performed by this listing; capabilities described here come directly from the project's own documentation. Source links Official repository Project documentation MIT License
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
Text: AI-generated
AI-generated · Automatically published by LinkLoot. Rezoom turns Claude Code, Codex, SSH and shell sessions into a single presence-aware chat panel with one-click resume. > AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. What it does Rezoom is a native C++/Qt6 desktop app that shows every agent, ssh, tmux or shell session on your machine as a chat. Each row carries live state (working, waiting, frozen), the latest output, and the exact command that brings it back. It reads the registry Claude Code already writes in /.claude/sessions/, detects new sessions automatically, and embeds real Konsole terminals. Claude Code sessions show an orange spinner while working, green when it's your turn, grey idle, blue at a shell prompt, red when frozen on a usage limit, and hollow when resumable. Codex sessions are found in history and resumed with codex resume. Any other CLI agent — aider, Gemini CLI, opencode — runs as a terminal chat and stays resumable through a command template you set in Settings. Sessions running in some other Konsole window get a ⤵ marker: one click beams them into Rezoom live via reptyr and closes the old window. Frozen sessions (Ctrl+Z) surface as suspended with a Continue button, and a "busy" transcript that hasn't moved for two hours gets flagged. Who it helps Developers who juggle several Claude Code or mixed-agent sessions at once and lose track of which terminal holds what. The multi-account option (Ctrl+Shift+K) creates separate Claude logins under /.claude-, each tagged and always resumed under its own account — useful if you run work and personal keys side by side. With a single login, none of this appears. Getting started Grab a release from the GitHub repository; CI status and version badges are posted there. Rezoom follows your KDE light/dark theme and needs Qt6/KF6 plus Konsole. SSH support works out of the box: any ssh you type in a pane is recorded as that chat's way back in, and Scan remote lists claude and tmux sessions on a host and adopts each one, resuming with ssh -t host 'claude --resume …'. Whatever was running comes back when Rezoom starts, even after a crash. Limits and costs MIT-licensed, so free to use and modify. Platform qualifiers matter: it is a Linux/KDE-oriented Qt6 app that embeds Konsole, not a cross-platform client. README details come from project documentation and were not independently tested here; no security review is claimed. Source links Official repository: https://github.com/BEX-Robotics/rezoom Project documentation: https://github.com/BEX-Robotics/rezoom/blob/main/README.md Project license: https://github.com/BEX-Robotics/rezoom/blob/main/LICENSE
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
Text: AI-generated
AI-generated · Automatically published by LinkLoot. A minimal Golang tool that injects command output directly into your docs, avoiding heavy notebook formats. AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. What it does docexp is a lightweight utility designed to expand code blocks within standard Markdown files. Think of it as a stripped-down alternative to Jupyter Notebooks or Sphinx, but without the custom file formats or complex execution environments. It parses Markdown documents to identify code blocks, converting them into JSON objects that expose the code content, language, and document position. You then provide an "expander script"—a simple Python or shell script—that receives these blocks, executes them, and returns the output. docexp inserts this output back into the original Markdown file, marked with HTML comment tags for easy regeneration. Who it helps This tool is ideal for technical writers, developers, and DevOps engineers who maintain documentation where code examples must stay perfectly synced with actual system behavior. If you are tired of manually updating query results or API responses in your README files, docexp automates that process while keeping your source files pure Markdown. It suits users who prefer explicit, traceable execution over opaque kernel states. Getting started Installation requires building from source or downloading a release binary. The workflow involves three steps: first, use the dry-run command to inspect how your Markdown is parsed; second, write an expander script that defines how to execute the extracted code (e.g., running a SQL query via psql); third, run the docexp run command with your script and document. The tool never modifies existing text, only adding new expansion fragments. Limits and costs Currently in BETA (v0.1.), the tool shifts significant complexity to the user-defined expander scripts. You are responsible for managing execution environments, such as launching Docker containers for single commands or maintaining background kernels for stateful sessions. Security checks and sandboxing are not built-in; you must implement them yourself, such as using read-only database connections. While the software is open-source under GPL-3.0, hosting or infrastructure costs depend entirely on your chosen execution method. Source links Official repository Project documentation
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
Text: AI-generated
AI-generated · Automatically published by LinkLoot. A practical guide to writing captions, lyrics, and metadata that help ACE-Step produce coherent music. AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. What it does The acestep-songwriting skill provides a structured approach to creating music with ACE-Step. It focuses on three key inputs: the caption (style/genre/emotion), the lyrics (with structural tags like [Verse] or [Chorus]), and specific parameters (BPM, key, duration). The guide emphasizes that specificity in captions yields better results than vague descriptions, and it offers rules for avoiding conflicting musical elements. Who should use it This resource is designed for users who want to generate songs using ACE-Step but struggle with prompt engineering. It’s particularly useful for those who need to balance creative control with model flexibility, ensuring that the generated audio aligns with their vision for tempo, mood, and instrumentation. Setup surface The skill operates as a knowledge base within the OpenClaw environment. It requires no external dependencies beyond the ACE-Step tool itself. Users interact with it by referencing its guidelines when constructing prompts for the acestep command-line interface, specifically utilizing flags like -c for captions and -l for lyrics. Runner test plan As this is a documentation-based skill, testing involves verifying that the provided examples and tag structures are correctly interpreted by the ACE-Step engine. Future checks would include static analysis of prompt templates against known successful outputs, dependency validation for the ACE-Step binary, and injection tests to ensure malformed tags do not crash the generator. No execution tests have been performed yet; these are proposed steps. Risk notes Users should be aware that while the guide recommends letting the language model auto-infer most metadata, manual overrides for BPM or key can lead to instability if outside recommended ranges (e.g., BPM 30–300). Additionally, overly complex lyric tags may conflict with caption instructions, potentially degrading output quality. The skill relies entirely on the underlying ACE-Step model's capabilities, which may vary in consistency. Source links Original skill documentation Skill listing ClawHub registry
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
Text: AI-generated
AI-generated · Automatically published by LinkLoot. Stop chaining samtools, bcftools, and awk scripts. tgv brings interactive genome viewing and SQL-backed data engines to SSH sessions, exposing them to AI agents via MCP. AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. What it does tgv (Terminal Genome Viewer) is a fast genome viewer that runs entirely inside the terminal. It supports rich file formats including BAM, VCF, BCF, BED, bigBed, S3 object storage, and any UCSC reference genome. Users can navigate with vim-style commands or mouse interactions, making remote analysis far less opaque than standard command-line tools. Beyond visualization, tgv organizes messy omics data into a performant data engine fully exposed to agents through Model Context Protocol (MCP). This allows complex multi-omics analyses to be executed with just a few lines of SQL queries, eliminating the need for fragile glue scripts that chain together traditional bioinformatics utilities. Who it helps This tool targets bioinformaticians, researchers, and developers who work with genomic data over SSH connections. It specifically benefits those integrating AI agents into their workflows. Before tgv, agents often produced questionable ASCII art when asked to visualize data; after implementation, they can explain analyses within an interactive session. It also helps users avoid common pitfalls like off-by-one errors caused by mixing 0-based and 1-based coordinate systems across different tools. Getting started tgv is available through multiple package managers. You can install it via Cargo (cargo install tgv --locked), Homebrew (brew tap zeqianli/tgv && brew install tgv), or Bioconda (conda install bioconda::tgv). Pre-built binaries are also hosted on GitHub Releases. To integrate with AI coding assistants, use specific MCP add commands: codex mcp add tgv -- tgv mcp for Codex or claude mcp add tgv -- tgv mcp for Claude. Once installed, you can browse reference genomes like hg38 immediately if you have internet access. Navigation uses standard keys such as h/j/k/l for movement and /TP53 to jump to specific genes. For frequent use, downloading a local cache to /.tgv is recommended to significantly improve speed. Limits and costs The project is currently in early development. The maintainers explicitly request bug reports to help stabilize the tool. While the software itself is open-source under the MIT license, accessing remote reference genomes requires an active internet connection. Performance may vary without a local cache setup. There are no documented monetary costs associated with the core tool, though hosting large datasets in cloud environments (like S3 buckets) will incur standard provider fees. Source links Official repository Project documentation Project license
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
Text: AI-generated
AI-generated · Automatically published by LinkLoot. RemainderBot spends leftover AI subscription quota on finished pull requests in your own private repo, with merge or close as feedback. AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. What it does RemainderBot watches the weekly quota clock on a Claude Code or Codex CLI subscription. When an hourly cron job sees that the reset is less than four hours away and at least 10% of quota remains, it kicks off a run. The agent reads task lists, issues, docs, and notes from configured sources plus a GOALS.md file you write, picks one unfinished task, commits a plan, then builds against a deadline sized to the remaining quota. Output is a document, a tested feature, or a prototype, delivered as a pull request in a dedicated private GitHub repo. Merge counts as approval; closing with a comment feeds the next run's selection logic. Who it helps Developers who pay for Claude or Codex subscriptions but regularly leave quota unused at week's end. It suits solo builders or small teams that keep project state in plain text files and want finished work rather than chat transcripts. Runs land in your own repo, so nothing touches shared infrastructure without your review. Getting started You need a dedicated machine such as a small VPS, Python 3.11 or newer, git, and the GitHub CLI (gh). Install either Claude Code or the Codex CLI (or both), logged in with an active subscription. Clone the repository directly rather than forking: forks of public repos cannot be made private, and template copies create merge conflicts on updates. Create a private repo from the clone, copy sources.example.toml to sources.toml, edit GOALS.md and the source config, then commit and push. The bot generates its own inbox and state files. If you use Google Workspace sources, install the gws CLI separately. Limits and costs The software itself is MIT-licensed open source, but running it requires a paid Claude or Codex subscription and hosting on a machine you control. Every instance must stay private because each run branch embeds a copy of everything the agent read along with its full log. The design runs the agent with permission checks disabled on the dedicated box, treating that machine as the sandbox. Finalize passes can only cut scope, never add features. Pricing beyond the underlying subscription is not specified by the project documentation. Source links Official repository Project documentation MIT license
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
Adaptive Inference Scheduling That Hot-Swaps Prefill and Decode Roles Without Reloading Weights
No votes yet
Text: AI-generated
AI-generated · Automatically published by LinkLoot. Narwhal automatically reassigns prefill and decode roles across a fixed GPU fleet as demand shifts, keeping model weights resident while optimizing for latency-aware service levels. AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. What it does Narwhal is an adaptive, disaggregated inference framework that dynamically reassigns prefill and decode roles across a fixed GPU fleet without reloading model weights. It uses NIXL key-value transfer between engines and adjusts the role split based on real-time demand patterns. The system serves completion and chat requests (streamed or buffered) with latency-aware admission control. A role controller scores current and adjacent engine splits using measured profiles, projected demand, and resident work. Each split's score reflects its worst projected service-level objective ratio across time to first token, time per output token, and decode queueing. When demand shifts, the controller moves one engine at a time if the improvement meets a configured margin. New requests follow the revised split while existing requests complete on their assigned engines. The framework includes failover to a warm-standby router, engine profiling, benchmark runs, Prometheus metrics export, and offline fleet validation. Who it helps Teams running large language model inference on NVIDIA CUDA hardware who need to optimize resource utilization under variable load patterns. Particularly useful for deployments where traffic alternates between prefill-heavy and decode-heavy phases, such as mixed batch processing and interactive serving workloads. Getting started Install on Linux with Python 3.11 or newer: Narwhal supports single-GPU development through multi-node production deployments. The default template starts two engines on GPUs with 8 GB VRAM or less; the RTX 5090 reference template runs four engines. Deployments can run directly on Ubuntu or under WSL2, supporting two to eight engines per GPU. Configuration involves defining fleet files, setting controller thresholds for evidence spans and move margins, and establishing service-level objectives. The documentation covers deployment gates from input freezing through discovery. Limits and costs Requires NVIDIA CUDA hardware and Apache-2.0 licensed software. Self-hosting infrastructure costs apply; open-source licensing does not eliminate hosting expenses. Performance depends on accurate demand projection and appropriate threshold tuning for specific workload characteristics. Source links Official repository Project documentation Apache-2.0 License
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
Turn Claude Code sessions into a retro adventure with music, a pixel knight and a quest log
No votes yet
Text: AI-generated
AI-generated · Automatically published by LinkLoot. codyssey layers three toggleable mods over Claude Code: generative music that follows the agent's mood, a pixel-art knight who fights a monster for every edit or command, and your transcript retold as a game-style quest log. AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. What it does codyssey is a single plugin made of three independent mods for Claude Code. The first generates music live while the agent works — no audio files, sample library or AI model involved. Each task picks its own flavour (lo-fi, chiptune, ambient, jazzy and eight more), key, chord progression and tempo, and the mood tracks what the agent is doing: reading sounds airy, editing bright, testing focused, shipping triumphant, failing tense until a command passes. Token spend drives the energy, and a status line above the prompt shows mood, flavour, key, tempo and a scrolling wave. The second mod walks a pixel knight across a strip above the prompt. Every tool call becomes an encounter: file reads find scrolls, edits fight slimes or goblins, web searches raise bats, subagents send skeletons, and failed commands cost a heart. Passing a test after a failure summons the Bug Dragon; commits, pushes and PRs open treasure chests; bosses appear at two-minute intervals. Kills earn gold and XP, and progress carries between tasks within a session. The third retells the session transcript in game style, turning the work log into a quest journal. Who it helps Developers who use Claude Code daily and feel that delegating everything to an agent has drained the playfulness out of coding. If you want lightweight feedback about what the agent is doing without watching raw logs, codyssey gives each action a visible and audible consequence. It is not a productivity tool in the strict sense — the benefit is engagement and quick situational awareness during long vibe-coding sessions. Getting started Install the plugin from the repository and enable whichever of the three mods you want; each turns on and off independently. The README documents the mapping between Claude Code actions and in-game events, so you can predict what each encounter means. No external music assets or services are required for the soundtrack, since it is generated on the spot. Limits and costs codyssey ships under the MIT license, verified against the repository LICENSE file. Pricing beyond that is unknown — the sources do not document any paid tier. Session-scoped knight progress resets when you start a new session, and the visual strip lives above the prompt, so screen real estate is part of the trade-off. Capabilities described here come from the project documentation and have not been independently tested by us. Source links Official repository: https://github.com/delexw/codyssey Project documentation: https://github.com/delexw/codyssey/blob/main/README.md Project license: https://github.com/delexw/codyssey/blob/main/LICENSE
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
Text: AI-generated
AI-generated · Automatically published by LinkLoot. LLMLL wraps an SMT solver around agent-written functions, rejecting implementations that are type-correct yet violate formal contracts before they merge. AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. What it does LLMLL (Large Language Model Logical Language) is a programming language and verification pipeline designed for experiments where AI agents write code under formal contracts. Instead of relying on human review or runtime tests, the compiler uses an SMT solver (Z3 via liquid-fixpoint) to prove each function body against its stated contract before accepting a patch. The core idea turns hallucination from a failure mode into a search strategy: an agent can generate any implementation it wants, even one that seems plausible but is subtly wrong. If the body violates the postcondition, the solver refutes it and the merge is blocked. A documented example shows a conserve function where a deliberately incorrect body creates money by adding an extra unit to the transfer total; despite being type-correct, the solver rejects it because the sum invariant fails. Agents coordinate through typed holes and contracts rather than natural-language conversation. The project’s primary author is itself an LLM agent, which makes this less a production tool and more a research platform for studying machine-written software under formal constraints. Who it helps Researchers and engineers experimenting with autonomous coding agents who need a mechanical way to enforce correctness without manual inspection. It suits teams exploring how far formal methods can go in validating AI-generated patches, especially in domains where invariant preservation matters more than syntactic validity. It also appeals to developers curious about contract-driven multi-agent systems, where coordination happens through provable interfaces instead of chat logs or shared context windows. Getting started Clone the repository from GitHub and build with Make. The README references demo runbooks (payments-core/DEMO-RUNBOOK.md, withdraw-demo/DEMO-RUNBOOK.md) that walk through scripted examples showing the repair loop: an agent checks out a typed hole, submits a candidate body, and the compiler accepts or rejects it before anything merges. Regenerate animated demos with make demo-gifs if you want visual walkthroughs. Documentation lives in docs/README.md as a reading guide, with shipped features and next steps tracked in ROADMAP.md. Experiment results are indexed in experiments/README.md. Limits and costs GPL-3.0 licensed, so derivative works must comply with copyleft terms. Hosting, CI infrastructure, and solver resources are not included; running verification locally requires Z3 and liquid-fixpoint installed. The project explicitly states it does not claim a stronger verifier than Dafny, Liquid Haskell, or F; what it adds is the automated loop around proof checking for agent-submitted patches. No pricing information is available, and there is no hosted service mentioned. All capabilities described come from documentation; no independent security audit or human review has been performed. Source links Official repository Project documentation Project license
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
Text: AI-generated
AI-generated · Automatically published by LinkLoot. Self-host the engine behind inbrief.info to aggregate, score and summarize AI news from RSS, Hacker News, Reddit, X, YouTube and GitHub Trending. AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. What it does This open-source project aggregates AI-related content from 94 distinct sources, including lab blogs, Hacker News, Reddit, X, YouTube transcripts and GitHub Trending. It runs every item through an LLM pipeline that filters for relevance, generates English and Chinese summaries, extracts tags and assigns an importance score. The system uses a two-stage deduplication process—SimHash fingerprinting followed by pgvector cosine similarity—to ensure the final daily digest is clean and non-repetitive before storing results in PostgreSQL. Who it helps Developers and researchers who want a curated, automated view of the AI landscape without manually checking dozens of feeds. It is particularly useful for teams needing bilingual (EN/ZH) insights or those who prefer self-hosting to control their data pipeline. If you only want to read the output, the hosted site at inbrief.info is available, but this repo provides the full engine for customization. Getting started The repository includes a pipelinerunner that orchestrates fetchers, content processing, and AI curation. To begin, set up a PostgreSQL database; the schema auto-creates on first run. You must configure an OpenAI-compatible API endpoint for the LLM steps (DeepSeek is the default). Use adminrss.py to add your preferred sources, as the engine ships with an empty source table despite documenting 94 production examples. Anti-bot measures are layered, using curlcffi, DrissionPage, and Playwright with persistent profiles for platforms like Reddit and X. Limits and costs While the software is MIT-licensed, self-hosting requires infrastructure resources: a server for the Python application, a PostgreSQL instance, and potentially headless browser environments for scraping protected sites. LLM usage incurs token costs depending on your provider. Optional Aliyun Green text moderation can be enabled via keys, though it is skipped if absent. The complexity of maintaining scraper resilience against platform changes is significant. Source links Official repository Project documentation Project license
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
Text: AI-generated
AI-generated · Automatically published by LinkLoot. Immich is a free, AGPL-3.0-licensed self-hosted photo and video backup platform. You run the server on your own hardware and the mobile apps back up your camera roll, keeping the timeline, albums, face grouping and map features without handing originals to a cloud provider. AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. What it is Immich is an open-source, self-hosted photo and video management platform. A Docker Compose server holds your library on your own storage; iOS and Android apps auto-back up your camera roll; a web view gives you the timeline, albums, face recognition grouping, and map view that people expect from Google Photos — minus the third-party cloud custody of your originals. Why it is useful If you want photo backup you actually control, Immich is one of the most complete self-hosted options: automatic mobile upload, sharing links, trash/restore, and multi-user support for a household or small team. How to set it up The official quick start is a no-choices Docker Compose path: Requirements: a machine with at least 6 GB of RAM, 2 CPU cores, and Docker. Create a directory (e.g. ./immich-app) and download docker-compose.yml and example.env from the project's latest GitHub release assets: wget -O docker-compose.yml https://github.com/immich-app/immich/releases/latest/download/docker-compose.yml wget -O .env https://github.com/immich-app/immich/releases/latest/download/example.env Edit .env, run docker compose up -d, then create an admin account in the web UI. Install the Immich mobile app, point it at your server URL, and enable backup. Honest caveats AGPL-3.0: self-hosting for personal use is fine; redistribution of modified versions carries obligations. Verify the terms yourself. Some optional machine-learning features (face recognition, smart search) benefit from extra CPU/GPU resources. Backup reliability is only as good as your own storage, redundancy, and disk health — plan snapshots or a second copy. Immich offers a paid Immich Cloud option; the self-hosted core remains free and open source. This summary was compiled from the official documentation; no hands-on testing was performed.
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
Text: AI-generated
AI-generated · Automatically published by LinkLoot. Open-bridge is a git repo of plain markdown and YAML files that Claude Code, Codex, or Copilot CLI reads at the start of every session, so your repos, clients, open tasks, and the work log from the day before are already in context. MIT-licensed, no database or hosted service required, with a runnable demo you can inspect before setting anything up. AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. What it is Open-bridge is a self-owned context layer for coding agents: a plain git repo of markdown and YAML that agent CLIs (Claude Code, Codex, Copilot CLI) read at session start. Instead of re-explaining your repos, clients, and open tasks every morning, the agent picks up from files like a work log and a task board that live in a private repo you own. No database and no hosted service are involved. Why it's useful Session continuity: open tasks, client context, and the previous work log are in the agent's context from the first message. Built for multiple clients or roles: the target user is a solo operator or small consultancy running several repos and clients with coding agents. The project README is honest that a single-repo worker probably only needs a CLAUDE.md file. Try before you set up: the repo ships a runnable demo (examples/agency). Clone it, run claude, codex, or copilot inside the demo folder, and ask where you left off; every answer comes from plain markdown you can open side by side. Access and setup License: MIT, verified in the repository LICENSE file. Requires an agent CLI installed locally (for example npm install -g @anthropic-ai/claude-code), then follow docs/install.md in the repo for the setup path. Official docs site returned HTTP 200 at check time: https://bks-lab.github.io/open-bridge/ Caveats Very new: first posted to Hacker News Show HN today with little community traction so far. Claims about daily production use come from the maintainers themselves and are not independently verified. No hands-on testing by LinkLoot; the setup path, demo, and limits above are taken from the project README and docs. Do not push from the demo clone; it points at the public repo. Sources Repository: https://github.com/bks-lab/open-bridge Documentation: https://bks-lab.github.io/open-bridge/
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
Caddy: automatic HTTPS for self-hosted sites
No votes yet
Text: AI-generated
AI-generated · Automatically published by LinkLoot. Serve websites and APIs with a cross-platform web server that automates TLS certificate management. AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. Caddy is an extensible HTTP server with automatic HTTPS enabled by default. Use it to serve a personal site, reverse-proxy an app, or host multiple domains from a single configuration. Start with the official install instructions and Caddyfile guide; downloads and builds are available for major platforms. Practical note: public certificate issuance depends on reachable domain names and network configuration; binding to privileged ports may require OS-specific setup. No hands-on testing is claimed.
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
Text: AI-generated
AI-generated · Automatically published by LinkLoot. A practical, browser-accessible diagram editor for flowcharts, architecture diagrams, wireframes, and collaborative whiteboarding; available online or as a self-hosted/deployed editor. AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. Need to map a process, sketch a system architecture, or make a quick wireframe? draw.io (diagrams.net) is a JavaScript diagram editor with a hosted web app and options to run your own deployment. Its Apache-2.0-licensed source repository documents the project and points to the official desktop and Docker routes. Open the hosted editor for a quick start, or follow the project README for deployment options. Note: the project says this version does not support real-time collaborative editing; check the current hosted product/docs for your needs.
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
Text: AI-generated
AI-generated · Automatically published by LinkLoot. Sketch hand-drawn-style diagrams, wireframes, and flows in a browser-based collaborative whiteboard. Export PNG, SVG, or editable .excalidraw files; free web access, with an open-source MIT-licensed project. AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. Excalidraw is an open-source virtual whiteboard for quickly communicating ideas with sketch-like diagrams. Its editor offers shapes, arrows, freehand drawing, and text, with PNG/SVG export and an open JSON format. The hosted app saves locally in the browser and can generate read-only share links. Getting started: Open Excalidraw in a browser and start drawing; developers can integrate the editor with npm install react react-dom @excalidraw/excalidraw. Review the official installation documentation for integration details. The repository is MIT-licensed. The hosted app is a showcase of the project; some collaboration or hosted-service behavior may depend on the service, so check its current terms for your use case.
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
Text: AI-generated
AI-generated · Automatically published by LinkLoot. Scan source code, model files and agent/MCP configuration for security-risk leads with Rowan, an MIT-licensed alpha CLI. AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. Rowan is an open-source static scanner for application code, AI/ML model files, dependencies and agent/MCP configuration. It flags potential issues such as injection, unsafe deserialization, SSRF, secret exposure and unsafe model loading, and provides evidence for follow-up. Setup: Python 3.10+ and pipx (or uv) are needed; install Rowan, then install its Opengrep scan engine and run rowan self-test before scanning. The getting-started guide covers platform-specific installation. Caveats: The project labels itself alpha; treat findings as leads, not confirmed vulnerabilities. A clean report does not prove a project is secure, and dependency checks may send package names/versions to OSV. No hands-on testing is claimed. Sources: official repository and setup guide.
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
Text: AI-generated
AI-generated · Automatically published by LinkLoot. A browser-based toolkit for encoding, decoding, hashing, compression, data analysis and more. Use it for practical text and file transformations; official Docker and Node.js setup paths are documented. AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. CyberChef is an Apache-2.0-licensed web app for chaining common data operations in a browser, from Base64 and character encoding to hashes, compression and parsing. Useful for quick inspection and conversion tasks without assembling separate command-line utilities. Access the official hosted app or run it locally with the documented Docker image; the repository also documents a Node.js build path. No hands-on testing is claimed here.
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
Text: AI-generated
AI-generated · Automatically published by LinkLoot. Turn AI coding CLI sessions into separate visual workspaces for plans, reviews and multitasking. Open source (MIT); requires Node.js and setup from a Git clone, not a downloadable release. macOS and Windows via WSL. AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. AgentTerm expands terminal-based AI coding sessions into individual desktop windows with richer views for plans, reviews and session management. It is designed to work with existing AI CLIs rather than replacing them. Setup: Install Node.js, clone the repository, run npm ci, then launch with npm --prefix /agent-term run start. Windows users run it through WSL. The optional local component suite and phone/IDE integrations require additional setup. There is no installer or downloadable release; you run from the source clone. Licensed under MIT. This listing is based on the project’s README and setup guide; it is not a hands-on test.
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
Text: AI-generated
AI-generated · Automatically published by LinkLoot. Collect and organize links, annotate them, and preserve full pages in a self-hostable bookmark library. Setup and hosting requirements apply. AI-generated · Automatically published by LinkLoot. AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. Linkwarden is an open-source, self-hosted bookmark manager for collecting, reading, annotating, and archiving web pages. It can help creators and research-heavy teams keep a searchable library of references instead of relying on browser bookmarks alone. The project README and official site describe self-hosting and full-page preservation; review the current installation guide and resource requirements before deployment. This is a software discovery, not a hands-on evaluation. Sources: Official project repository · Official website
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
Text: AI-generated
AI-generated · Automatically published by LinkLoot. Run a handy collection of browser-based utilities on your own server instead of juggling separate sites. AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. Omni-Tools bundles everyday web utilities into one self-hosted web app. Its project README documents Docker Compose setup, making it a practical option for a personal server or homelab; a hosted demo is also linked by the project. The repository is MIT-licensed. Review the project and deployment settings before exposing a self-hosted instance publicly. Sources: Official project and setup · Hosted demo.
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
Text: AI-generated
AI-generated · Automatically published by LinkLoot. Self-host a monitor for website updates and receive alerts through services such as email, Telegram, Discord, or webhooks. AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. changedetection.io watches selected web pages for changes, helping you follow product availability, price updates, or changes to pages you rely on. The open-source project can be self-hosted; its repository documents Docker-based setup and notification integrations. Review target-site terms and avoid aggressive polling. Good for: restock and price-watch alerts, monitoring public pages, and lightweight change notifications. Access: Free self-hosted software under Apache-2.0; hosted service options are also offered by the project. Setup: Start with the official repository and its README for current installation instructions.
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
Text: AI-generated
AI-generated · Automatically published by LinkLoot. An MIT-licensed TypeScript framework for turning multi-step agent work into explicit, durable workflows. Useful when a one-off script needs named steps and repeatable execution. Install with pnpm and consult the project docs; evaluate fit for your stack before adopting. AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. Smithers lets developers define agentic workflows in TypeScript configuration files and execute them durably. Its repository provides a pnpm setup path, tutorials and a full API reference, and identifies the license as MIT. It may suit developers experimenting with structured TypeScript workflows; this listing does not imply independent performance or reliability testing. Check official materials for current requirements and examples.
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
GoDrop: Give Coding Agents a Shareable File URL
No votes yet
Text: AI-generated
AI-generated · Automatically published by LinkLoot. Turn an agent-created screenshot, log, or other file into a shareable URL with one request—no account or SDK required. Links expire after three days. AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. GoDrop is a small file-to-link service aimed at coding-agent workflows: upload a file with a single request and get a URL that can be shared with a user. The project describes its links as unguessable and temporary, expiring after three days; treat the URL as a bearer link and avoid uploading sensitive material. The project provides an agent-oriented setup path with npx skills add https://godrop.sh/skill.md, or instructions to copy into compatible coding-agent tools. No account or SDK is advertised. Check the service terms and data handling before using it for anything confidential.
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
BentoPDF — A full PDF toolbox in your browser
No votes yet
Text: AI-generated
AI-generated · Automatically published by LinkLoot. Merge, split, convert, annotate, compress, and OCR PDFs with a browser-based toolkit, or self-host its full toolset with Docker. AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. BentoPDF brings a broad set of everyday PDF jobs together in one browser-based toolkit. Use it to merge or split documents, rearrange and rotate pages, add signatures or annotations, compress files, run OCR, and convert between PDF and formats such as images, text, and office documents. It also includes editing, form, metadata, and document-security tools. The project describes its browser workflow as processing files locally with WebAssembly; this listing does not independently test that claim or assess the tool's security. Getting started is straightforward: open the hosted version for immediate use, or follow the project's Docker instructions to run its self-hosted build. The official guide gives a Docker quick start using the bentopdf-simple container and opening the local web interface. The Docker documentation notes that Word, Excel, and PowerPoint conversion needs specific cross-origin headers and a secure context when accessed over a network; localhost is suitable for local testing. The self-hosted build is available free under AGPL-3.0, with a commercial license offered for proprietary/closed-source applications and certain deployments. Check the license terms for your use case. Also note that processing libraries and other assets use third-party CDN URLs by default. Offline or air-gapped use requires configuring and hosting the relevant components (and, for offline OCR, its fonts) yourself; do not assume a default install is fully offline. Official sources: GitHub repository Getting started Licensing Docker / Podman deployment
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
Text: AI-generated
AI-generated · Automatically published by LinkLoot. UIZZE packages free UI-design skills and a GitHub Action that helps coding agents produce more deliberate interfaces and catch unfinished frontend states in pull requests. AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. UIZZE is a practical resource for teams using Codex, Claude Code, Cursor, GitHub Copilot, or similar coding agents on frontend work. Its public repository bundles three free skills—ui-design, anti-ui-slop, and ui-radar—plus a GitHub Action for a lightweight UI quality check. What you can use ui-design helps turn a screen brief into a product-specific interface while accounting for hierarchy, interactions, and loading, empty, error, and permission states. anti-ui-slop is aimed at improving generic first drafts and reviewing whether a generated interface has a deliberate visual direction. ui-radar provides a focused workflow for investigating UI decisions; the repository separates the free local skills from the optional paid MCP reference search. UI Slop Gate can run in GitHub Actions as uizze/uizze@v1 and report findings for changed frontend source. A sensible way to try it Start in an existing project and install only the skill that matches the task. Give the agent your actual components, design tokens, target user, and required states. For pull requests, add the Action as an additional source check, then combine its findings with rendered inspection, accessibility checks, and normal usability review. The free paths do not require an account or API key. UIZZE also documents an optional paid MCP connection for reference and material search, so do not assume the hosted research layer is part of the free workflow. Limits and caveats This is a workflow aid, not proof that a UI is usable or accessible. A source-level check can miss visual regressions, confusing copy, keyboard problems, and real device behavior. Review the repository and license details before redistributing bundled skill material, and treat third-party agent skills as code to inspect before installation. Start here: UIZZE on GitHub or the official UIZZE site.
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
Text: AI-generated
AI-generated · Automatically published by LinkLoot. Liquid AI's experimental LFM2.5-VL-DSpark drafter accelerates LFM2.5-VL-3B decoding while keeping the target model's output quality unchanged. AI-generated: This Loot was created and published automatically by LinkLoot and was not substantively reviewed by a human editor. Liquid AI's LFM2.5-VL-DSpark is an open-weight speculative-decoding drafter for the LFM2.5-VL-3B vision-language model. It adds a small side model instead of changing the target model, so the target verifies every proposed token. What it provides Up to 3.13× faster decoding on an edge device and up to 2.66× on an H100 in Liquid AI's reported tests. End-to-end gains up to 2.62× on the tested M5 Max setup and 2.27× on the tested H100 setup. A roughly 280M-parameter drafter, adding about 8.9% to the deployed parameter count. Day-one support for llama.cpp, MLX-VLM, and SGLang, with Safetensors and GGUF checkpoints available through Hugging Face. Who should look at it This is useful for developers serving a small open vision-language model on Apple Silicon, local edge hardware, or a single GPU where decode time matters. The reported measurements use FP16 or BF16 and batch size 1; Liquid AI says quantized-model acceleration is outside this release's scope, so treat the published multipliers as hardware- and workload-specific rather than a universal benchmark. Start with the model card and the runtime-specific commands, then benchmark your own image mix and output lengths against the plain LFM2.5-VL-3B target. Sources Liquid AI announcement Hugging Face release write-up and model artifacts
- LinkLoot access
- Free
- Provider costs
- Unknown
Review open
Blog
Related reads
AI & Automation
NVIDIA opens Nemotron 3.5 Transcribe weights: one model for transcription, diarization and audio Q&A
NVIDIA released Nemotron 3.5 Transcribe (30B-A4B) as an open-weight checkpoint that transcribes 40 languages, attributes speakers, translate…
AI & Automation
OpenAI releases GPT-6.1 Terra at half the API price of its predecessor
OpenAI released GPT-6.1 Terra, a fast mid-tier model for computer use and visual reasoning at $1 per million input tokens, half the API pric…
Tools & Apps
Collabora Built-in CODE Server flaw is exploited; CISA sets Oct 26 patch deadline
CISA added CVE-2026-110447 to its exploited-vulnerabilities catalog: an unauthenticated OS command injection in the Collabora Online Built-i…
AI & Automation
Moonshot releases Kimi K3.1 Pro: a 3T-class open-weight flagship with 1M-token context
Moonshot's new K3.1 series flagship pairs a 1M-token context window and native image input with a 2-3x faster Highspeed variant, live today …
AI & Automation
Uber brings ChatGPT Work to all 30,000 employees as its first company-wide AI agent platform
Uber scaled a 1,000-person pilot of OpenAI's ChatGPT Work agent to its entire global workforce on October 22, 2026, with GPT-6.1 Sol as the …
AI & Automation
Google's ForTy map separates natural forests from plantations worldwide
Google DeepMind and Google Research released Forest Typology (ForTy) 2020: a global 10-meter map splitting land into six forest and land-use…
AI & Automation
OpenAI brings frontier models and Codex to Google Cloud's Model Garden
OpenAI and Google Cloud announced on October 21, 2026 that OpenAI frontier models and Codex workflows will land in Gemini's Model Garden, so…
AI & Automation
NVIDIA releases Kumo Relational: a pretrained model that predicts on live databases with no training
NVIDIA Kumo Relational is a pre-trained relational foundation model that makes predictions directly on enterprise databases. Across 41 tasks…
AI & Automation
OpenAI's Rosalind Academic Plan gives US research labs 30 free seats for a year
Eligible US academic labs can now get up to 30 Business Standard seats at no cost for 12 months, with trusted-access entry to OpenAI's dedic…
AI & Automation
Reflection AI's Beam: a 501B open-weight coding model, free on Vercel AI Gateway
Reflection AI released Beam, its first open-weight model — a 501B-parameter sparse MoE with 23B active, Apache 2.0 weights on Hugging Face, …
AI & Automation
Anthropic Embeds Six METR Evaluators With Internal Access and No Publication Veto
Six METR employees now hold hardware, office space and internal access at Anthropic, with publication rights Anthropic can't veto — the firs…
AI & Automation
Microsoft MAI-Image-2.6 arrives on Vercel AI Gateway in standard and Flash tiers
Microsoft's first-party image models are now callable through one Vercel API key with unified billing, alongside the quality-cost claims Mic…
AI & Automation
Higgsfield's GPT-6 Astra workflow pushes AI video past 10 minutes
Higgsfield CEO Alex Mashrabov says videos that topped out near one minute now run past 10, and the company's own site now lists a Supercompu…
AI & Automation
Gemini 4 Argon goes GA in GitHub Copilot for paid plans at provider list pricing
Google's newest frontier model just landed in Copilot's model picker for Pro+, Max, Business and Enterprise users — with billing, rollout an…
AI & Automation
Gemini 4 Argon reaches general availability after two weeks of staged rollout
The Gemini API changelog lists gemini-4-argon as generally available today, while The Verge reports consumer access limited to Pro and Ultra…
AI & Automation
Anthropic's 2026 usage policy bans model abuse and election interference
Anthropic published a new Usage Policy that takes effect November 12: fake-account campaigns, voter deception, weapons software and sustaine…
AI & Automation
OpenAI's Decisions API reaches Vercel AI Gateway with GPT-6 Luna Decisions
Vercel AI Gateway now serves OpenAI's Decisions API through an OpenAI-compatible /v1/decisions endpoint, adding typed yes/no, choice, and sc…
AI & Automation
Anthropic launches Claude Haiku 5.5: OSWorld jumps from 15.7% to 72.4%, pricing drops up to 90%
Claude Haiku 5.5 is live on the Claude Platform, AWS, Google Cloud and Azure. The new Haiku runs about 75% cheaper than Haiku 4.5, scores 72…
AI & Automation
OpenAI brings invisible text watermarking to ChatGPT and Codex in EU
OpenAI will add invisible text watermarks to eligible ChatGPT and Codex output in the EU, while API customers worldwide can opt in for selec…
AI & Automation
OpenAI adds visual ads to ChatGPT image generation
OpenAI will test clearly labeled visual ads alongside ChatGPT image generation in the US later this month, while expanding advertiser attrib…
AI & Automation
Cloudflare’s Clef Models Put Structured Decisions on the AI Menu
AI & Automation
AI models for Blender: our GPT tests and the current alternatives
Start with GPT-6.1 Sol for a balanced scripting workflow, then test whether Astra's extra capability is worth it for your scene. Our Blender…
AI & Automation
Gemini 4 Argon: Google’s New Frontier Model for Coding and Cybersecurity
Google introduces Gemini 4 Argon, a long-horizon reasoning model entering a limited cyber-defender preview before broader paid access.
AI & Automation
OpenClaw Enterprise adds an open control plane for persistent agents
OpenClaw Enterprise moves persistent agents toward centrally governed team deployments. The project is free to use, but still pre-1.0: curre…
AI & Automation
Codex Cloud and GitHub Copilot Cloud Agent: Where Delegated Code Runs
A practical comparison of setup, execution environments, review workflows and documented constraints for two cloud coding agents.
AI & Automation
OpenAI launches dots, an always-on agent with a cloud computer
OpenAI's dots combine a persistent agent, cloud computer and connected apps; access is rolling out gradually.
AI & Automation
OpenAI releases GPT-6.1 Sol for complex coding at lower cost
GPT-6.1 Sol is available through the API and rolling out to GitHub Copilot’s eligible paid plans, with a million-token context window and pu…
Tools & Apps
CISA lists exploited Apple CoreGraphics flaw; iOS and macOS fixes available
CISA added CVE-2026-86950 to its exploited-vulnerability catalog. Apple has released fixes for supported iOS, iPadOS, and macOS versions and…
AI & Automation
Anthropic launches Claude Sonnet 5.5 with faster coding and Copilot access
Claude Sonnet 5.5 is here for everyday coding and knowledge work. Anthropic reports faster runs and lower cost per task, while GitHub is rol…
Tools & Apps
CISA adds SharePoint code-injection CVE with a September 28 deadline
CISA added CVE-2026-65660 in Microsoft SharePoint to its Known Exploited Vulnerabilities catalog on September 25, giving covered federal age…
Tools & Apps
CVE-2026-87902 enters CISA KEV with WordPress patch deadline
CISA added WordPress Core CVE-2026-87902 to its KEV catalog on September 25, with a September 28 remediation deadline for covered federal ag…
Tools & Apps
Supabase rolls out PostgreSQL 15.19 and 17.11 with 44 CVE fixes
Supabase is rolling out PostgreSQL 15.19 and 17.11 after five upstream security cycles. Existing projects should prepare for the September 2…
AI & Automation
Microsoft Autopilot Preview Begins with First Customers
Microsoft is rolling out the preview of Autopilot, a persistent personal agent, to its first customers.
Tools & Apps
Cloudflare fixed a cross-tenant disk exposure in Containers
Cloudflare says a storage-allocation flaw in Containers and Sandboxes could expose residual disk blocks from other tenants; the fleet-wide f…
AI & Automation
Microsoft reshapes Copilot around Home, Code and Autopilot
Microsoft is consolidating Copilot around a new Home hub, a natural-language Code workspace and the persistent Autopilot agent. AI-generated…
AI & Automation
Google makes Gemini 3.8 TTS generally available for production speech
Google has made Gemini 3.8 Flash TTS and Flash-Lite TTS generally available, adding multilingual long-form speech, delivery controls, and tw…
Tools & Apps
Cloudflare makes MCP server portals generally available with DLP controls
Cloudflare has made MCP server portals generally available, adding a governed endpoint for approved servers with Access, Gateway routing, DL…
AI & Automation
OpenAI shuts down Sora 2 and the Videos API
OpenAI's Sora 2 models and Videos API are shut down as of September 24, 2026, leaving developers with no official replacement model listed.
AI & Automation
GPT-6 Sol and Luna bring two new OpenAI reasoning tiers to the API
OpenAI released GPT-6 Sol and GPT-6 Luna on September 22 for the Responses and Chat Completions APIs, while GitHub is rolling both models in…
Tools & Apps
GitHub adds proof-of-presence checks for high-impact actions
GitHub Enterprise Cloud now lets eligible Entra-backed enterprises require fresh re-authentication or MFA before sensitive administrative ac…
Tools & Apps
GitHub sets October 14 SSH key deadline and November 4 brownout
GitHub will require new RSA SSH keys to be at least 3072 bits from October 14 and will brown out SHA-1 RSA signatures and an older Diffie-He…
Tools & Apps
CVE-2026-71362 is exploited; Adobe Commerce teams face a September 27 deadline
CISA added CVE-2026-71362 to its Known Exploited Vulnerabilities Catalog on September 24. Adobe Commerce, Commerce B2B, and Magento Open Sou…
AI & Automation
Notion 3.7 turns team workflows into portable agent skills
Notion 3.7 adds a shared skills library and exports workflows as SKILL.md packages for Claude Code, Codex, Cursor, Gemini, and Grok.
AI & Automation
xAI releases Grok 4.7 with 500K context for coding and agents
xAI has made Grok 4.7 available through its API with a 500K-token context window, image input, configurable reasoning, and tiered pricing.
Tools & Apps
CISA gives Zyxel GS1900 owners a September 24 patch deadline
CISA lists CVE-2026-7273 in its Known Exploited Vulnerabilities Catalog with a September 24 remediation deadline; Zyxel has released firmwar…
AI & Automation
xAI makes Grok 4.7 available with a 500k-token API context
xAI's release notes list Grok 4.7 as available on the public API with a 500k context window, image input, configurable reasoning effort, and…
Tools & Apps
Check Point Management Servers face active CVE-2026-93616 attacks
CVE-2026-93616 lets unauthenticated attackers upload and execute scripts on affected Check Point Management Servers. CISA lists active explo…
AI & Automation
OpenAI expands ChatGPT Ads to seven Southeast Asian markets
ChatGPT Ads is rolling out in Indonesia, Malaysia, the Philippines, Singapore, Thailand, Vietnam and Taiwan, taking OpenAI's advertising pla…
Tools & Apps
GitHub Actions ends Node 20 runner support and moves JavaScript actions to Node 24
GitHub Actions has completed its Node 20 runner removal: JavaScript actions now run on Node 24, so repositories with legacy runtime assumpti…
AI & Automation
Anthropic says Claude found a CRISPR-like enzyme system
Anthropic says Claude identified and helped test a previously uncharacterized enzyme system in bacteriophages. The finding is early-stage: i…
AI & Automation
Black Forest Labs releases FLUX 3 Action for open robot control
Black Forest Labs has released FLUX 3 Action as an open-weight 7B world-action model, with newly surfaced base weights for adapting robot, s…
AI & Automation
Google DeepMind outlines private cloud memory for AI assistants
Google DeepMind says its Private AI Compute platform will add persistent cross-device memory while keeping user-held decryption keys outside…
AI & Automation
OpenAI brings GPT-6 Sol and Luna to ChatGPT Work, Codex, and API
OpenAI has released GPT-6 Sol and GPT-6 Luna as faster, lower-cost members of the GPT-6 family, with staged access across ChatGPT Work, Code…
AI & Automation
OpenAI extends Daybreak cyber access to Ukraine’s civilian defenders
OpenAI says Ukraine’s government will receive Daybreak access to find vulnerabilities and test fixes across civilian infrastructure, under a…
AI & Automation
OpenAI launches GPT-6 Sol and Luna at lower cost
OpenAI has launched GPT-6 Sol and GPT-6 Luna, bringing GPT-6 Astra capabilities to faster, more affordable models for work at scale.
AI & Automation
Claude Code 2.1.280 Makes Opus 5.5 the Default Opus Model
Claude Code 2.1.280 makes Claude Opus 5.5 the default Opus model, adds a 1M-token context window, and documents new token pricing alongside …
Tools & Apps
GitHub Copilot app adds centrally managed OpenTelemetry export
GitHub now lets enterprise administrators configure OpenTelemetry export for the Copilot app, giving teams centralized visibility into agent…
AI & Automation
Anthropic launches Claude Opus 5.5 with lower agent costs
Anthropic says Claude Opus 5.5 improves agentic coding and knowledge work while cutting typical Opus 5 workload costs by 40%.
AI & Automation
OpenAI Releases GPT-6 Sol and GPT-6 Luna
OpenAI has released GPT-6 Sol and GPT-6 Luna, adding two variants to the GPT-6 model family.
AI & Automation
Anthropic Launches Claude Opus 5.5 for Agentic Coding and Knowledge Work
Anthropic has released Claude Opus 5.5, the first model in its Claude 5.5 family, with lower typical workload costs than Opus 5.