🦁 A research-friendly codebase for fast experimentation of multi-agent reinforcement learning in JAX

🏛️A research-friendly codebase for fast experimentation of single-agent reinforcement learning in JAX • End-to-End JAX RL

Set up and improve harness engineering (AGENTS.md, docs/, lint rules, eval systems, project-level prompt engineering) for AI-agent-friendly codebases. Triggers on: new/empty project setup for AI agents, AGENTS.md or CLAUDE.md creation, harness engineering questions, making agents work better on a codebase.

Makes your codebase agent-friendly and keeps it that way. Small files so agents don't waste tokens. CLAUDE.md in every directory so they don't have to guess. Tests so they check their own work. Run once, it bolts rules into your global config — every project enforces standards automatically.

Audit a codebase for AI-agent friendliness — diagnose friction patterns, score with AFC Score, and deliver a prioritized improvement roadmap.

Discover, classify, and inventory AI agents in your codebase—deterministically, privately, and CI-friendly.

agent友好型代码仓库构建skill

Conventions that make codebases searchable, predictable, and agent-friendly

MCP server that generates AI-agent-friendly directory maps for better codebase navigation

PortPilot is an agent-friendly CLI (and MCP server) for porting a codebase **incrementally, commit-by-commit**.

AI-powered agent that analyzes codebases, detects quality issues, and generates actionable developer-friendly reports with interactive Q&A.

🧬 Rust-powered codebase intelligence engine — reveals tech stack, architecture, LOC distribution, dead code, and more. Agent-friendly JSON output.

a small, config-driven toolkit for measuring how agent-friendly a codebase is without relying on an LLM to infer the result.

Pardon_Me is an AI Agents GitHub repository interpreter that turns complex codebases into clear, conversational explanations. It analyzes user intent, fetches live repo data, detects jargon, and delivers beginner-friendly insights like a teammate who already understands the entire project.

A codebase that enables websites to be **agentic-browser friendly**. BotSight crawls a webpage, validates the scrape, extracts structured data, generates metadata/snippets, and ships it as an npm module/snippet that site owners can embed. The goal is to make pages visible and actionable for AI agents and LLM crawlers.

Surveyor is a developer tool that scans your codebase and produces a structured, AI-friendly report of your project. It extracts layered context (repo → files → symbols), compresses it into a deterministic snapshot, and enables AI agents to reason about code without loading the entire repository into context.