"Done" means nothing — a menhera-style completion gate for Claude Code that blocks empty completion claims until tests actually ran green and no TODOs are left behind.
Make AI coding agents prove they're done. A fail-closed completion gate: the agent only proposes done; CI reads the real artifacts and grants 'complete'. For Claude Code, Codex & any agent.
Evidence gate for AI agents: NO COMPLETION CLAIMS WITHOUT FRESH VERIFICATION EVIDENCE. Cross-cutting skill for TriaDev / Claude Code workflow.
Evidence rungs and a completion gate: an AI cannot claim done without the evidence.
kkamak completion gate — Claude Code cannot say done until your check passes
Multi-agent frugality defense plugin for DeepSeek Harness: read-ledger dedup, compaction-immune rules, completion gate, cheap-review lane.
Verifiable completion for DeepSeek Harness: the plugin runs your real project checks, pins results to the working-tree state, expires stale evidence, and gates turn completion until proof exists. | 插件自己执行验收命令、把结果钉在工作区指纹上、证据随代码变化自动过期，收尾时先取证再宣称完成。
To associate your repository with the completion-gate topic, visit your repo's landing page and select "manage topics."