lesson-capture

Lesson Capture

Record structured lessons after significant tasks. Builds a persistent knowledge base that improves future agent performance. Run with an optional summary or use review to read past lessons.

Trigger

User invokes /lesson-capture [SUMMARY] or says "capture lesson", "what did we learn", "record this"

Examples:

/lesson-capture Fixed the content pipeline to handle missing images gracefully
/lesson-capture The blog QA caught 3 canon violations — updated banned phrases
/lesson-capture review

Input: A brief summary of what happened, or review to read recent lessons. Output: A formatted lesson entry appended to the monthly lessons file, or a pattern analysis of recent lessons.

---

When to Capture

Record a lesson after:

• Multi-phase tasks — pipelines, proposals, audits (what phase was hardest? what would you skip?)
• Debugging sessions — root cause analysis (what was the actual problem vs what you initially suspected?)
• User corrections — any time the user said "no, not that" or "actually, do it this way" (what assumption was wrong?)
• Non-obvious discoveries — workarounds, undocumented behaviors, surprising file locations
• Failed approaches — something you tried that did not work (save others from the same mistake)
• Validated approaches — something non-obvious that worked well (reinforce the pattern)

Do not capture: routine tasks, trivial fixes, or lessons already documented in existing commands.

---

Auto-Lesson on QA Failure

When /lesson-capture is invoked after a QA failure (content-qa, blog-qa, or any validation step that returned failures), automatically detect and capture the failure pattern:

Detection

Check the current session for recent QA results by looking for:

• Files matching {{CONTENT_OUTPUT_DIR}}/**/qa-report*.md
• Recent conversation mentions of "FAIL", "violation", "rejected", "not compliant"
• Output from /content-qa or /blog-qa within this session

Auto-Capture Format