update-skills

You can analyze session transcripts to extract successes, failures, missing capabilities, and redundancies. This conversation review identifies what went wrong in multi-agent orchestration and categorizes issues into new-skill needs, scope conflicts, and workflow gaps.

A postmortem review for agent skills examines completed sessions to learn from failures and prevent recurring issues. You need this conversation review after multi-step sessions where agents miss conventions or produce conflicting guidance, ensuring skills evolve continuously.

To update SKILL.md files, generate prioritized proposals from session findings and interactively apply targeted edits to files and related indexes. This process presents diffs for your review, accepting approvals before implementing step-by-step skill updates.

Yes, you can prioritize proposed new skills and workflow updates by classifying extracted findings from agent dialogues. This categorization process ranks recommended changes by severity, separating new-skill needs, updates, conflicts, and scope issues for targeted skill governance.

Yes, this approach works for multi-agent orchestration breakdowns by analyzing the session to identify conflicts, missing capabilities, and vagueness. It categorizes these workflow gaps and produces implementable skill updates to prevent similar orchestration failures.