Milestone retrospective — analyze historical .planning/ data for process improvement insights. Run after milestone completion to reflect on estimation accuracy, verification patterns, and common gaps. **Process learning through quantitative analysis** Retrospectives are not blame sessions. They identify systemic patterns that improve future planning: - Were estimates accurate? (planned phases vs actual including gap closures) - Did verification catch issues? (first-try pass rate) - What gaps recur? (common patterns across phases) Call pan-tools retro to gather all metrics: ```bash RETRO=$(node ~/.claude/pan-wizard-core/bin/pan-tools.cjs retro) ``` Parse JSON for: `phases_planned`, `phases_completed`, `phases_decimal`, `estimation_accuracy_pct`, `verifications_total`, `verifications_passed_first_try`, `verifications_gaps_found`, `verifications_human_needed`, `first_try_rate_pct`, `common_gap_patterns`. **If error:** Report "No roadmap found — run /pan:new-project first." **Estimation Accuracy Analysis** ``` Planned phases: {phases_planned} Completed: {phases_completed} Gap closure phases: {phases_decimal} Estimation accuracy: {estimation_accuracy_pct}% ``` | Accuracy | Assessment | Recommendation | |----------|-----------|----------------| | ≥ 90% | Excellent estimation | Keep current approach | | 70-89% | Good, some gaps expected | Minor calibration needed | | 50-69% | Under-estimated scope | Add buffer phases or break work smaller | | < 50% | Significant under-estimation | Review requirements gathering process | **Verification Pattern Analysis** ``` Verifications: {verifications_total} Passed first try: {verifications_passed_first_try} ({first_try_rate_pct}%) Gaps found: {verifications_gaps_found} Human needed: {verifications_human_needed} ``` | First-try rate | Assessment | Recommendation | |---------------|-----------|----------------| | ≥ 80% | Strong execution quality | Current workflow is effective | | 60-79% | Moderate gap rate | Improve plan detail or must_haves quality | | < 60% | High gap rate | Review plan-checker effectiveness, add more explicit must_haves | **Common Gap Patterns** If `common_gap_patterns` is non-empty: ``` Top recurring gap types: 1. {pattern} — appeared {count} times 2. {pattern} — appeared {count} times ... ``` For each top pattern, suggest a preventive action: - Missing wiring → "Add explicit key_links to must_haves" - Stub detection → "Strengthen plan-checker substance checks" - Test failures → "Add test gate earlier in verification" - Missing files → "Require artifact existence in plan verify steps" **Retrospective Report** ```markdown ## Milestone Retrospective ### Estimation - Phases: {completed}/{planned} ({decimal} gap closures) - Accuracy: {accuracy}% — {assessment} - Recommendation: {recommendation} ### Verification Quality - First-try pass rate: {rate}% — {assessment} - Gap phases needed: {gaps_found} - Human verification: {human_needed} - Recommendation: {recommendation} ### Common Gap Patterns {pattern_list_with_recommendations} ### Key Takeaways 1. {takeaway based on data} 2. {takeaway based on data} 3. {takeaway based on data} ### Next Milestone Adjustments - {specific adjustment for next milestone planning} ``` - Read-only: this workflow does not modify any files - Data-driven: all assessments backed by quantitative metrics - Actionable: every finding includes a concrete recommendation - Non-judgmental: focus on systemic patterns, not individual decisions