lessons-harvester

Automating lessons learned capture involves converting human interventions from Jira epics and Confluence comments into structured lessons. This process applies evidence-based inputs and strict human-in-the-loop validation before persisting entries to a lessons.md file.

You should run a lessons harvester after closing an iteration, merging PRs, or closing Jira epics. It triggers during retrospectives or lessons requests to capture evidence-based interventions and convert them into structured, reusable lessons.

No, the lessons harvester appends new lessons to an existing lessons.md file without overwriting previous entries. It maintains sequential numbering using atomic lesson identifiers (L-XXX) and groups them by process, technical, execution, or integration categories.

Generating structured lessons requires evidence-based inputs such as specs with multiple versions, Confluence MCP access, and Jira data. These inputs provide the necessary context for drafting lessons before human-in-the-loop validation persists them.

Yes, automating retrospective lessons generates a decisions log for traceability. It groups validated lessons by category—process, technical, execution, or integration—ensuring each captured intervention links back to its originating context.

Human-in-the-loop validation for automated lessons capture ensures accuracy before persisting drafted lessons to the file. It prevents incorrect or unverified pipeline interventions from becoming permanent entries in your structured lessons.md repository.