review-mistakes

Review pending AI mistake reports by interactively triaging SKELETON-*.md and MISTAKE-*.md files. The workflow summarizes each report and guides you through Accept, Reject, Discard, Enrich, or Skip decisions to harden AI workflows.

Accepted mistake reports are integrated by appending corresponding entries to AGENTS.md. The reviewed report files are then organized into dedicated integrated, rejected, or discarded directories to maintain a clear audit trail.

Find pending MISTAKE and SKELETON markdown files by searching the project's .intent-layer/mistakes/pending directory. The review session automatically lists these files to begin the interactive triage process.

You can enrich a mistake report during the interactive triage process. Enrich is one of the five available decision options alongside Accept, Reject, Discard, and Skip, allowing you to add context before finalizing the report's status.

The intent layer mistake review process requires no external dependencies. It operates entirely within your project environment to automate the triage and integration of mistake reports into your operational workflow.

Triage SKELETON markdown files to identify and integrate structural learning loop insights into AGENTS.md. This continuous review prevents recurring AI workflow pitfalls by maintaining an updated operational audit trail.