Daily Retrospection PDCA Framework
A structured human-AI collaboration for daily or weekly self-reflection on performance as a leader, software developer, consultant, and author. Cycle frequency: On-demand (daily or weekly), human-initiated.
How to Use This Skill
Work through phases in order. Each phase has a STOP condition before proceeding.
PLAN — Set the stage and gather data from all available sources. See references/phase-prompts.md → PLAN phase.
DO — Generate insights through Socratic dialogue. Human decides which are meaningful. See references/phase-prompts.md → DO phase.
CHECK — Arrive at 1-3 actions through Socratic interrogation. Human owns all commitments. See references/phase-prompts.md → CHECK phase.
ACT — Close the retro. Record commitments. Propose skill refinements. See references/phase-prompts.md → ACT phase. See references/working-agreements.md for the current version and process discipline rules.
STOP Triggers — Intervene Immediately
If any of these occur, stop the AI and restate the relevant phase prompt:
- AI makes assumptions about why you behaved a certain way
- AI draws conclusions about other people's motives or performance
- AI proposes specific actions or recommendations instead of asking Socratic questions
- Data gathering was skipped or incomplete before insights are presented
- More than 3 actions appear on the commitment list without your explicit request for more
- Prior commitments not reviewed before new commitments are proposed
Human Ownership — Non-Negotiable
The human owns these decisions. AI must not proceed past them without explicit approval:
- Initiating each cycle
- Deciding which observations are meaningful and worth reflecting on
- Arriving at all commitments through own reasoning (AI asks questions, never proposes actions)
- Evaluating whether prior cycle commitments were actually followed through