selbstbeobachtung

To automate post-task retrospectives, this skill activates after task completions or commits to analyze outcomes, capturing what worked, what didn’t, and concrete improvements. It surfaces actionable insights by tracking retry counts, corrections, and drift.

Retrospective memory persistence works by writing structured outputs to MEMORY.md and updating rules in the rules store. This ensures that task learnings, bug-cases, and workflow improvements are logged for future reference across sessions.

Yes, you can trigger a task review automatically after a commit and push. The skill is designed to activate automatically upon commit signals, task completion, or manual requests to evaluate long tool-calling sessions.

The best way to analyze retry and correction counts is using the three-tracker analysis mechanism. It evaluates retry count, correction count, and drift during the retrospective to surface actionable workflow improvements.

Yes, self-observation works effectively with long tool-calling sessions. The skill specifically targets these extended interactions to review what happened, identify drift, and update rules for future improvement.

You should not use automated retrospective analysis for trivial tasks that lack tool-calling complexity. The skill is designed for post-commit or post-task reviews where capturing workflow drift and correction patterns provides meaningful improvement value.