retrospective

Analyzing completed AI agent sessions involves ingesting session logs and state files to detect inefficiencies, quality issues, and process gaps, then generating actionable improvement suggestions for skills and agent prompts.

An AI factory retrospective is a post-completion analysis of agent interactions that identifies knowledge gaps and AGENTS.md health issues, needed after complex software development sessions to pinpoint delays and propose concrete fixes.

Updating AGENTS.md based on agent performance requires analyzing session artifacts to identify process flaws, then generating specific AGENTS.md update proposals that require human approval before application.

Automating suggestions for new AI agent skills is possible by analyzing session logs and project documentation to identify efficiency and quality patterns, proposing new or modified skills for human review before implementation.

Retrospective analysis does not automatically apply changes to agent prompts; it generates concrete improvement suggestions for skills, prompts, and AGENTS.md updates that require explicit human approval before application.

The best way to review AI agent artifacts for process gaps is conducting pattern analysis across efficiency, quality, and knowledge dimensions using session logs and state files to generate targeted improvement proposals.