Agent Friendly Code: Which public repos are friendliest to an AI coding agent?

Hey PH 👋 Maker here. Which public repos are friendliest to an AI coding agent? I kept asking this after watching the same agent one-shot a task in one repo and flail for an hour in another. The variable was never the agent — it was the repo. So I built the leaderboard that answers it. 🎯 Per-model — Claude Code weights AGENTS.md, Devin cares about CI, Cursor likes strong types. "Friendliest" changes with the agent you use. 🔍 Explainable — every repo page shows the top 3 fixes to climb, ranked by points gained. A to-do list, not just a grade. 🌐 Cross-forge — GitHub, GitLab, Bitbucket out of the box. Plus a [PR Action](https://github.com/hsnice16/agen...) that comments your score-delta on every commit, and a [local skill](https://github.com/hsnice16/agen...) that scores your repo and picks a model. Honest caveat: per-model weights come from each agent's docs today — running every agent on every repo (the real benchmark) is the 1.0 goal. Shipping the useful 80% now, upfront about the rest. What signal matters most for your agent that I'm not measuring yet? That shapes the rubric directly. 🙏