/mcs-retro — Post-Session Retrospective

You can capture and classify development session learnings by analyzing build errors, eval failures, and manual workarounds. This updates local markdown files and index.json, integrating observations into a structured knowledge system to prevent knowledge loss.

The best way to classify build errors and workarounds is to categorize them as REPEAT, NEW, CORRECTION, ENHANCEMENT, TOOLING_GAP, SOLUTION_PATTERN, or DECISION_OUTCOME. This intelligent classification cross-references against a solution library to enrich the data for continuous improvement.

Automated retrospective classification integrates with knowledge management by updating local markdown learning files and index.json. It cross-references session observations against a solution library and cache to ensure enriched classification and structured knowledge system integration.

Yes, you can use this retrospective tool to identify tooling gaps from agent build sessions. It classifies items under the TOOLING_GAP category and suggests improvements based on the manual workarounds and build errors encountered during the development session.

Yes, the continuous improvement classification system depends on the @google/generative-ai library. This dependency provides the underlying reasoning required to analyze build session learnings and classify them into structured categories like SOLUTION_PATTERN or DECISION_OUTCOME.

You should run a post-session retrospective immediately after a complex agent build session that encountered unexpected errors or required manual workarounds. This logs issues, classifies them, and records solution patterns and decision outcomes for future reference.