retro

Agent memory analysis reads MEMORY.md files stored under .claude/agent-memory and aggregates repeated patterns to surface automation opportunities. It flags context bloat when line counts exceed thresholds and outputs a retrospective report with skill candidates and sub-agent proposals.

Agent retrospective reporting reviews accumulated agent memory to identify repeated patterns and context bloat. You need it when your MEMORY.md files grow large and you want to extract skill candidates, sub-agent proposals, and future design guidelines from past interactions.

Context bloat detection checks MEMORY.md files by comparing line counts against predefined thresholds. When lines exceed the threshold, the retrospective report flags it, helping you identify when agent memory has grown too large and needs optimization or skill extraction.

Yes, retrospective analysis aggregates repeated patterns found in MEMORY.md files and outputs sub-agent proposals. These proposals suggest new sub-agents to handle recurring tasks, reducing duplication and improving automation efficiency based on historical interaction data.

Yes, agent retrospectives require MEMORY.md files to be stored under the .claude/agent-memory directory. Without this directory structure and the associated memory files, the retrospective analysis cannot read interaction history or generate automation opportunity reports.