tacoda コミュニティtack-eval
Prove the repo's committed context actually makes agents effective — run canary tasks that a fresh agent must complete using only the text files (AGENTS.md, docs, ADRs), measure pass/fail, and feed every failure back into the context in the same PR. Load when you want evidence a brownfield repo is agent-ready, want to measure whether context quality improves over time, or want CI to catch when the text stops representing the truth. Part of the tack distribution.
- 1時間の導入数
- 1
- GitHub Stars
- 0
- 更新日
- 11月22日