agent-standards
Discovers your codebase's real coding conventions from evidence, then feeds only the relevant ones to your AI coding agent.
Build, learn & keep up — tech, Linux & sports news included.
86 guides, built from each repo's own README. Full step-by-step activation guides for every repo remain in the classic site; this page is the searchable index with each guide's summary and official link.
Discovers your codebase's real coding conventions from evidence, then feeds only the relevant ones to your AI coding agent.
A safety-first crypto trading bot for Python: paper trading by default, risk rules enforced in code, and a local AI that can only say no.
A fact-based market regime and technical-analysis desk: deterministic signals, risk vetoes in code, and paper mode unless you opt in by hand.
A zero-trust governance console that watches autonomous AI agents: catches prompt injection, credential leaks and rogue tool calls in real time.
A paper-mode skeleton that scans memecoins on Dexscreener with hard factual filters — and deliberately has no live trading path at all.
A curated architecture guide: the 8 open-source repos that make up a serious AI trading stack, organised layer by layer.
A big curated library for beginner bug bounty hunters: basics, setup, tools, labs and vulnerability guides — organised as a table of contents.
LinkToVid: a Telegram bot that turns a video or music link into a file in your chat — yt-dlp under the hood, Docker on top.
A ChatGPT-style chat interface you run yourself, with many model providers behind one login.
An AI agent from HKUDS that researches markets for you: pulls price data, reads the news, writes up the analysis.
A Bloomberg-style market terminal built by the community: quotes, charts, screeners and news in one dense screen.
A voice-driven AI character with a Live2D avatar that talks back — and can run fully offline on your own machine.
A stealth, Firefox-based browser that AI agents drive — handing them compact page snapshots instead of raw HTML soup.
By HeyGen. Write a video as HTML and CSS — timed scenes in a web page — and render it into a real MP4 file.
A skill that plugs into Claude Code and audits your Google and Meta ad accounts like a media buyer hunting waste.
A free, open-source, local-first video editor positioned as a CapCut replacement for macOS, Windows, Linux and Android.
A lightweight, agentic IDE built around the pi coding agent, from JetBrains, with git-worktree workspaces and living SPEC.md files.
A multi-agent LLM trading framework that mirrors a real trading firm, from analysts debating to a portfolio manager's final call.
A free, open-source SEO plugin for Claude Code that runs a full website audit in 10–15 minutes and returns a prioritised action plan.
A fast, open-source word processor written in pure Rust, presented as a clean-room reimplementation of Microsoft Word.
A single-prompt agent skill that makes an AI coding agent behave like a famously lazy senior developer: reuse first, write the minimum that works.
A small, personal collection of agent skills for shipping pull requests, cleaned up for anyone to use in Claude Code and Codex.
A high-performance coding-agent harness written in Rust, built for heavy multi-agent development workflows.
A plug-and-play research framework for 'recursive agent harnessing' in an open world — improving the system around a model, not the weights.
An independent auditor for AI agents that answers one question: is the agent actually doing what it is supposed to do?
An open-source personal AI assistant that runs on your own computer and meets you in the chat apps you already use.
A complete software development methodology for coding agents, built from composable skills that make an agent brainstorm, plan and build with discipline.
A collection of small, composable agent skills for real engineering work — requirements grilling, test-driven development, bug diagnosis and triage.
A menu-bar app and local gateway that lets every coding agent use any model — and quietly fails over when one account's quota runs out.
The open-source AI coding agent (formerly OpenCode) — a terminal-based agent with a full-access build agent and a read-only plan agent.
Turns a codebase — plus docs, SQL schemas, configs and PDFs — into a queryable knowledge graph you explore instead of grepping through files.
A free, open-source file converter you use by dragging: hold a key, drag files, and drop them on a format in a donut menu at your cursor — entirely offline.
A local-first search engine and MCP server for AI agents, shipped as a single Rust binary — index code, docs, logs and PDFs for full-text, vector and hybrid search.
Describe a test goal in plain language and an agent drives your web or mobile app to reach it — then later runs replay the recorded steps with no model calls.
A self-hostable alternative to Vercel, Heroku and Netlify — deploy apps, databases and 360+ one-click services on your own servers.
A policy-enforced runtime that lets autonomous agents use files, packages, APIs and credentials without giving them unrestricted access.
A full signing stack that marks AI-generated images in a publisher's own name, with content credentials, watermark and fingerprint.
An open-source, AI-native vector design tool that turns prompts into live canvas designs and exports them as code.
An offline-first desktop office suite in Rust, with documents, spreadsheets, presentations, PDF, mail, Git and Markdown apps that keep files in your own folders.
A fast, ad-free replacement home screen for Google TV, LG webOS and Samsung Tizen, with a one-button way back to the factory screen.
A sandbox platform for running agent and user code in isolated Firecracker microVMs, controlled over an HTTP API.
A fast, parallel programming language that lets you state app rules as laws and demand a mathematical proof that code obeys them.
A memory system that captures what your coding agent does, compresses it, and injects the relevant context back into future sessions.
A design skill for AI coding agents that replaces generic AI-looking interfaces with a shared design vocabulary and deterministic quality checks.
A lightweight, open-source design agent that turns a page description into a single HTML file, and a working example of a durable agent built on libfx.
Let your coding agent reverse-engineer software without source code — native binaries, Electron apps, .NET assemblies and websites — via MCP or a terminal CLI.
Turn an ad-filled Android TV / Google TV home screen into a calm Apple TV-style one over Wi-Fi — backup, telemetry and bloat disabled, tiny native launcher installed, fully reversible.
A coding-agent skill from Cloudflare that turns your agent into a structured security auditor — six phases, adversarial validation, machine-readable findings.
A Rust archive extractor (CLI + desktop app) that extracts huge archives in place — the archive shrinks as your files hatch, so you only need free space for the output.
Import .sql files directly into TypeScript with full type inference and zero codegen, powered by TypeScript 7.1's content mappers.
The NSA's open-source reverse engineering framework — disassembler, decompiler, graphing and scripting for a wide range of processors and executable formats.
A desktop AI orchestrator for running a fleet of coding agents side by side — Claude Code, Codex, OpenCode and more — each in its own isolated git worktree, with a mobile companion.
Run open models locally — Kimi, GLM, MiniMax, DeepSeek, gpt-oss, Qwen, Gemma and more — with complete privacy, from one command.
Open-source, fully-local voice studio — voice cloning, voice design, video dubbing, dictation, transcription and audiobook creation in 646 languages. The open ElevenLabs alternative.
A self-hosted gym and body-weight tracker you own — 5,600+ exercise library with animated demos, workout logging, muscle maps and PRs. No subscription, no ads, no telemetry.
A Mac menu-bar dictation app: press a shortcut, speak, and the transcript pastes into whatever app you're using.
A local-first AI agent that browses the web, edits files and runs approved shell commands — with its control loop and memory on your machine.
It really boots iOS 16.5's user space on Android hardware — but on exactly one phone model, as a time-limited research bench.
Scan a room with a LiDAR iPhone and an AI agent rebuilds it as an editable Three.js scene — iteratively, not in one click.
'Spot it. Save it.' — one click saves AI images, videos AND their prompts from 20+ sites into a local library your coding agents can search.
An editing instruction kit for Codex and Claude Code: 15 skills and a playbook that teach your agent transcript-driven video editing.
Pixel art as code: write a short .px.js file and get PNGs, sprite sheets, animated GIFs, Tiled maps and Aseprite files — plus a live viewer.
Draw a box on a map and download that slice of the real world — buildings, terrain, satellite imagery, Street View and 3D models — for LLMs to work with.
A 2.3 MB native Mac front-end for KDE Connect: your Android phone's notifications, texts, files and clipboard, on your Mac.
University research code for agentic video generation: one persistent world state behind a whole film, so scenes stay consistent across cuts.
DeepSeek's own production inference framework, open-sourced — for exactly one model, on datacentre GPUs you almost certainly don't have.
EAGLE-4 speculative decoding for SGLang, vLLM and TokenSpeed — one draft model that adapts its speculation depth to your batch size and hardware.
Your Claude/Codex usage limits, rendered as pictures and pushed to a ~$5 GeekMagic Wi-Fi clock — no firmware flashing, because it uses the clock's stock photo album.
Turns a structure video — grey-box render, depth map, skeleton, sketch — into a finished styled video in four sampling steps.
Type a motion, get a rigged character performing it — but the quality pass needs a paid fal.ai key, Windows + an NVIDIA GPU, and two academic model registrations.
A desktop CAD app where a coding agent writes parametric models as CadQuery Python — you describe the part, point, measure, iterate, then export for printing.
A ~$7 ESP32 board that lets an AI agent see and drive your iPhone — no jailbreak. The catch: the board is the smallest cost in the setup.
An ESP32 dashboard showing live Gran Turismo 7 telemetry — straight from the PS5 over Wi-Fi, no PC needed — or via SimHub USB for PC games.
Not a photo→3D AI model: an agent skill that hand-authors an editable Blender scene from one room photo, through 8 stages and 3 review gates.
A skill that explains any concept by inventing a physical machine whose operation IS the concept — then builds it as an interactive isometric figure.
Massively multilingual zero-shot text-to-speech with voice cloning and voice design, from the k2-fsa speech team — with weights, a demo and a paper.
A free Mac uninstaller that finds every file an app leaves behind, shows WHY each file matched — and refuses to touch your keys, mail and backups.
One image in, a relightable textured 3D asset out — real research, but a Linux + NVIDIA pipeline and a gated Meta model hide behind the demo.
A Hermes Agent plugin that makes context compaction reuse the server's prompt cache instead of re-reading everything — author's benchmark: ~80s down to ~16s.
Plays Windows games on Apple Silicon Macs by assembling Wine, DXMT, D3DMetal and DXVK for you — free, with an open per-game compatibility database.
186 tested GLSL functions, a bundler and a compile checker for your coding agent — solid building blocks, despite the 'compiles first try' hype.
Boots Linux on a PS5 via a jailbreak payload and a rest-mode resume trick — real, and bannered EXPERIMENTAL by its own README.
A framework that gets frontier vision models to draw their reasoning onto images as editable annotations — it ships no model of its own.
Motion design for coding agents: numeric easing/spring rules, a self-review loop that inspects its own renders frame by frame, and a sound engine in pure Python.
A standalone GTK4 voice-input app for Linux: press a key, speak, and the text lands in the focused window — no API key needed on the default path.
'Mission control for agent skills' — one fast terminal UI inventorying every skill across all your agents, flagging the broken ones.