conducting-post-mortem

You conduct a post-mortem by analyzing the current task conversation to extract actionable lessons, identifying up to three novel 'gotchas', and proposing concrete content updates to prevent repeating issues.

The best way to capture unknown unknowns is to audit the task conversation, prioritize up to three high-impact gotchas by rank, and generate concrete content proposals to address each one.

You update CLAUDE.md by generating guided content proposals from your task review, receiving precise placement guidance for where to insert the new lessons learned within your knowledge management file.

Yes, you can use a post-mortem review for end-of-task knowledge capture, as it automatically analyzes the session and optional task file to propose reusable patterns and updates without manual writing.

You should not use an automated post-mortem when a task lacks novel issues, because the skill prioritizes only up to three high-impact, previously unknown gotchas to avoid cluttering your knowledge base.