When you need to run multiple search passes, compare candidates across trials, and keep a visible audit trail of what won and why, this gives you the structure. It pushes you to log each rollout with timestamps, prior winners, search parameters, and explicit accept/watch/reject marks. The coherence check compares new results against previous ledgers so you can spot drift or staleness. Promotion rules are strict: paper mode by default, and you only go live when the new candidate beats the prior winner, passes replay checks, and gets explicit user approval. Built for trading models, config migrations, or any high stakes optimization where recursive confidence alone is not proof.
npx -y skills add affaan-m/everything-claude-code --skill recursive-decision-ledger --agent claude-codeInstalls into .claude/skills of the current project.
Select a file.
mattpocock/skills