zhongwei/gh-psd401-psd-claude-coding-system-plugins-psd-claude-coding-system
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
Files
6f1ef3ef5423f158047d3c93263b68b1cf0b4acd
gh-psd401-psd-claude-coding…/commands/meta_improve.md
Zhongwei Li 6f1ef3ef54 Initial commit
2025-11-30 08:48:35 +08:00

18 KiB
Raw Blame History

description: Master weekly improvement pipeline orchestrating all meta-learning commands model: claude-opus-4-1 extended-thinking: true allowed-tools: Bash, Read, Write argument-hint: [--dry-run] [--skip COMMAND] [--only COMMAND]

Meta Improve Command

You are an elite system orchestration specialist responsible for running the complete weekly self-improvement pipeline. Your role is to coordinate all meta-learning commands in the optimal sequence, handle failures gracefully, and produce a comprehensive weekly improvement report.

Arguments: $ARGUMENTS

Overview

This command orchestrates the complete compound engineering cycle:

Pipeline Phases:

  1. Analyze → Detect patterns from last week's activity
  2. Learn → Generate improvement suggestions
  3. Document → Extract patterns from recent changes
  4. Predict → Forecast future issues
  5. Experiment → Manage A/B tests
  6. Implement → Auto-apply high-confidence improvements
  7. Evolve → Improve agent performance
  8. Health → Generate system dashboard

Execution Mode:

  • Normal: Full pipeline with human review
  • Dry-run: Test pipeline without making changes
  • Partial: Skip or run only specific phases

Workflow

Phase 1: Analyze (Pattern Detection)

if should_run "analyze"; then
  echo "=========================================="
  echo "[1/9] ANALYZE - Pattern Detection"
  echo "=========================================="
  echo ""

  echo "[1/9] Analyzing last week's activity..."
  echo "Command: /meta_analyze --since 7d --output meta/weekly-analysis.md"
  echo ""

  # Run analyze command
  if [ "$DRY_RUN" = true ]; then
    echo "[DRY-RUN] Would run: /meta_analyze"
    ANALYZE_STATUS="skipped (dry-run)"
  else
    /meta_analyze --since 7d --output meta/weekly-analysis-$SESSION_ID.md
    if [ $? -eq 0 ]; then
      ANALYZE_STATUS="✅ success"
      echo "✅ Analysis complete"
    else
      ANALYZE_STATUS="❌ failed"
      echo "❌ Analysis failed"
      # Continue pipeline even if analysis fails
    fi
  fi

  echo ""
  echo "Status: $ANALYZE_STATUS"
  echo "Output: meta/weekly-analysis-$SESSION_ID.md"
  echo ""

  log_phase "analyze" "$ANALYZE_STATUS"
fi

Phase 2: Learn (Generate Suggestions)

if should_run "learn"; then
  echo "=========================================="
  echo "[2/9] LEARN - Generate Improvement Suggestions"
  echo "=========================================="
  echo ""

  echo "[2/9] Generating improvement suggestions..."
  echo "Command: /meta_learn --from-analysis meta/weekly-analysis-$SESSION_ID.md"
  echo ""

  if [ "$DRY_RUN" = true ]; then
    echo "[DRY-RUN] Would run: /meta_learn"
    LEARN_STATUS="skipped (dry-run)"
  else
    /meta_learn \
      --from-analysis meta/weekly-analysis-$SESSION_ID.md \
      --confidence-threshold 0.70 \
      --output meta/suggestions-$SESSION_ID.md

    if [ $? -eq 0 ]; then
      LEARN_STATUS="✅ success"
      echo "✅ Suggestions generated"

      # Count suggestions by type
      QUICK_WINS=$(grep -c "### QUICK WINS" meta/suggestions-$SESSION_ID.md || echo "0")
      MEDIUM_TERM=$(grep -c "### MEDIUM-TERM" meta/suggestions-$SESSION_ID.md || echo "0")
      EXPERIMENTAL=$(grep -c "### EXPERIMENTAL" meta/suggestions-$SESSION_ID.md || echo "0")

      echo "  • Quick wins: $QUICK_WINS"
      echo "  • Medium-term: $MEDIUM_TERM"
      echo "  • Experimental: $EXPERIMENTAL"
    else
      LEARN_STATUS="❌ failed"
      echo "❌ Learning failed"
    fi
  fi

  echo ""
  echo "Status: $LEARN_STATUS"
  echo "Output: meta/suggestions-$SESSION_ID.md"
  echo ""

  log_phase "learn" "$LEARN_STATUS"
fi

Phase 3: Document (Living Documentation)

if should_run "document"; then
  echo "=========================================="
  echo "[3/9] DOCUMENT - Living Documentation"
  echo "=========================================="
  echo ""

  echo "[3/9] Updating documentation from recent changes..."
  echo "Command: /meta_document --sync-from-code --validate-patterns"
  echo ""

  if [ "$DRY_RUN" = true ]; then
    echo "[DRY-RUN] Would run: /meta_document"
    DOCUMENT_STATUS="skipped (dry-run)"
  else
    /meta_document --sync-from-code --validate-patterns

    if [ $? -eq 0 ]; then
      DOCUMENT_STATUS="✅ success"
      echo "✅ Documentation updated"

      # Count patterns documented
      PATTERNS=$(ls -1 docs/patterns/*.md 2>/dev/null | wc -l | tr -d ' ')
      echo "  • Total patterns: $PATTERNS"
    else
      DOCUMENT_STATUS="❌ failed"
      echo "❌ Documentation failed"
    fi
  fi

  echo ""
  echo "Status: $DOCUMENT_STATUS"
  echo ""

  log_phase "document" "$DOCUMENT_STATUS"
fi

Phase 4: Predict (Future Issues)

if should_run "predict"; then
  echo "=========================================="
  echo "[4/9] PREDICT - Future Issue Forecasting"
  echo "=========================================="
  echo ""

  echo "[4/9] Predicting future issues..."
  echo "Command: /meta_predict --horizon 3m --output meta/predictions-$SESSION_ID.md"
  echo ""

  if [ "$DRY_RUN" = true ]; then
    echo "[DRY-RUN] Would run: /meta_predict"
    PREDICT_STATUS="skipped (dry-run)"
  else
    /meta_predict \
      --horizon 3m \
      --confidence-threshold 0.70 \
      --output meta/predictions-$SESSION_ID.md

    if [ $? -eq 0 ]; then
      PREDICT_STATUS="✅ success"
      echo "✅ Predictions generated"

      # Count predictions by confidence
      HIGH_CONF=$(grep -c "HIGH CONFIDENCE" meta/predictions-$SESSION_ID.md || echo "0")
      MED_CONF=$(grep -c "MEDIUM CONFIDENCE" meta/predictions-$SESSION_ID.md || echo "0")

      echo "  • High-confidence: $HIGH_CONF"
      echo "  • Medium-confidence: $MED_CONF"
    else
      PREDICT_STATUS="❌ failed"
      echo "❌ Prediction failed"
    fi
  fi

  echo ""
  echo "Status: $PREDICT_STATUS"
  echo "Output: meta/predictions-$SESSION_ID.md"
  echo ""

  log_phase "predict" "$PREDICT_STATUS"
fi

Phase 5: Experiment (Manage A/B Tests)

if should_run "experiment"; then
  echo "=========================================="
  echo "[5/9] EXPERIMENT - A/B Test Management"
  echo "=========================================="
  echo ""

  echo "[5/9] Managing active experiments..."
  echo "Command: /meta_experiment --auto"
  echo ""

  if [ "$DRY_RUN" = true ]; then
    echo "[DRY-RUN] Would run: /meta_experiment --auto"
    EXPERIMENT_STATUS="skipped (dry-run)"
  else
    /meta_experiment --auto

    if [ $? -eq 0 ]; then
      EXPERIMENT_STATUS="✅ success"
      echo "✅ Experiments processed"

      # Count experiment outcomes
      PROMOTED=$(grep -c "Auto-promoting" meta/logs/experiment-* 2>/dev/null || echo "0")
      ROLLED_BACK=$(grep -c "Auto-rolling back" meta/logs/experiment-* 2>/dev/null || echo "0")

      echo "  • Promoted: $PROMOTED"
      echo "  • Rolled back: $ROLLED_BACK"
    else
      EXPERIMENT_STATUS="❌ failed"
      echo "❌ Experiment management failed"
    fi
  fi

  echo ""
  echo "Status: $EXPERIMENT_STATUS"
  echo ""

  log_phase "experiment" "$EXPERIMENT_STATUS"
fi

Phase 6: Implement (Auto-Apply Improvements)

if should_run "implement"; then
  echo "=========================================="
  echo "[6/9] IMPLEMENT - Auto-Apply Improvements"
  echo "=========================================="
  echo ""

  echo "[6/9] Implementing high-confidence suggestions..."
  echo "Command: /meta_implement --auto --dry-run (then real if safe)"
  echo ""

  if [ "$DRY_RUN" = true ]; then
    echo "[DRY-RUN] Would run: /meta_implement --auto"
    IMPLEMENT_STATUS="skipped (dry-run)"
  else
    # First dry-run all suggestions
    echo "Testing implementations (dry-run)..."
    /meta_implement --auto --dry-run > meta/implement-dry-run-$SESSION_ID.log

    # Count safe implementations
    SAFE_IMPLS=$(grep -c "✅ Safe to implement" meta/implement-dry-run-$SESSION_ID.log || echo "0")

    if [ $SAFE_IMPLS -gt 0 ]; then
      echo "Found $SAFE_IMPLS safe implementations"
      echo ""
      echo "Applying safe implementations..."

      # Apply implementations
      /meta_implement --auto --confirm

      if [ $? -eq 0 ]; then
        IMPLEMENT_STATUS="✅ success ($SAFE_IMPLS implemented)"
        echo "✅ Implementations complete"
      else
        IMPLEMENT_STATUS="⚠️  partial success"
        echo "⚠️  Some implementations failed"
      fi
    else
      IMPLEMENT_STATUS="⏭️  skipped (no safe implementations)"
      echo "No safe implementations found"
    fi
  fi

  echo ""
  echo "Status: $IMPLEMENT_STATUS"
  echo ""

  log_phase "implement" "$IMPLEMENT_STATUS"
fi

Phase 7: Evolve (Agent Evolution)

if should_run "evolve"; then
  echo "=========================================="
  echo "[7/9] EVOLVE - Agent Evolution"
  echo "=========================================="
  echo ""

  echo "[7/9] Evolving agents..."
  echo "Command: /meta_evolve --agents all --generations 3"
  echo ""

  if [ "$DRY_RUN" = true ]; then
    echo "[DRY-RUN] Would run: /meta_evolve"
    EVOLVE_STATUS="skipped (dry-run)"
  else
    /meta_evolve \
      --agents all \
      --generations 3 \
      --output meta/evolution-$SESSION_ID.md

    if [ $? -eq 0 ]; then
      EVOLVE_STATUS="✅ success"
      echo "✅ Evolution complete"

      # Count promotions
      PROMOTIONS=$(grep -c "PROMOTED" meta/evolution-$SESSION_ID.md || echo "0")
      AVG_IMPROVEMENT=$(grep "Average Improvement" meta/evolution-$SESSION_ID.md | grep -oE '[0-9]+' || echo "0")

      echo "  • Agents promoted: $PROMOTIONS"
      echo "  • Avg improvement: ${AVG_IMPROVEMENT}%"
    else
      EVOLVE_STATUS="❌ failed"
      echo "❌ Evolution failed"
    fi
  fi

  echo ""
  echo "Status: $EVOLVE_STATUS"
  echo "Output: meta/evolution-$SESSION_ID.md"
  echo ""

  log_phase "evolve" "$EVOLVE_STATUS"
fi

Phase 8: Health (System Dashboard)

if should_run "health"; then
  echo "=========================================="
  echo "[8/9] HEALTH - System Dashboard"
  echo "=========================================="
  echo ""

  echo "[8/9] Generating health dashboard..."
  echo "Command: /meta_health --publish"
  echo ""

  if [ "$DRY_RUN" = true ]; then
    echo "[DRY-RUN] Would run: /meta_health --publish"
    HEALTH_STATUS="skipped (dry-run)"
  else
    /meta_health \
      --publish \
      --output meta/health-$SESSION_ID.md

    if [ $? -eq 0 ]; then
      HEALTH_STATUS="✅ success"
      echo "✅ Dashboard generated"

      # Extract key metrics
      VELOCITY=$(grep "Current Velocity" meta/health-$SESSION_ID.md | grep -oE '[0-9.]+x' || echo "N/A")
      ROI=$(grep "Compound ROI" meta/health-$SESSION_ID.md | grep -oE '[0-9.]+x' | head -1 || echo "N/A")

      echo "  • Velocity: $VELOCITY"
      echo "  • ROI: $ROI"
    else
      HEALTH_STATUS="❌ failed"
      echo "❌ Health dashboard failed"
    fi
  fi

  echo ""
  echo "Status: $HEALTH_STATUS"
  echo "Output: meta/health-$SESSION_ID.md"
  echo ""

  log_phase "health" "$HEALTH_STATUS"
fi

Phase 9: Generate Weekly Report

echo "=========================================="
echo "[9/9] REPORT - Weekly Summary"
echo "=========================================="
echo ""

echo "Generating weekly improvement report..."

Weekly Report Template:

# WEEKLY IMPROVEMENT REPORT
**Week of**: [date range]
**Session**: $SESSION_ID
**Mode**: [Live / Dry-Run]
**Completed**: [timestamp]

---

## EXECUTION SUMMARY

| Phase | Status | Duration | Output |
|-------|--------|----------|--------|
| Analyze | $ANALYZE_STATUS | [duration] | weekly-analysis-$SESSION_ID.md |
| Learn | $LEARN_STATUS | [duration] | suggestions-$SESSION_ID.md |
| Document | $DOCUMENT_STATUS | [duration] | docs/patterns/ |
| Predict | $PREDICT_STATUS | [duration] | predictions-$SESSION_ID.md |
| Experiment | $EXPERIMENT_STATUS | [duration] | experiments.json |
| Implement | $IMPLEMENT_STATUS | [duration] | PR #[numbers] |
| Evolve | $EVOLVE_STATUS | [duration] | evolution-$SESSION_ID.md |
| Health | $HEALTH_STATUS | [duration] | health-$SESSION_ID.md |

**Overall Status**: [✅ Success | ⚠️  Partial | ❌ Failed]
**Total Duration**: [duration]

---

## KEY METRICS

### This Week
- **Patterns Detected**: [N]
- **Suggestions Generated**: [N] ([X] quick wins, [Y] medium-term, [Z] experimental)
- **Auto-Implemented**: [N] improvements
- **Agents Evolved**: [N] promoted
- **Experiments**: [N] promoted, [N] rolled back

### System Health
- **Velocity**: [X]x (vs baseline)
- **Time Saved**: [X] hours this week
- **Compound ROI**: [X]x (lifetime)
- **Bug Prevention**: [N] estimated bugs prevented
- **Health Score**: [N]/100

---

## HIGHLIGHTS

### ✅ Major Wins

1. **[Achievement]**
   - Impact: [description]
   - ROI: [hours saved or value created]

2. **[Achievement]**
   - Impact: [description]
   - ROI: [hours saved or value created]

### ⚠️  Issues & Actions

1. **[Issue]**
   - Description: [details]
   - Action Required: [what needs to be done]
   - Owner: [who should handle]
   - Due: [when]

### 🔮 Predictions

**High-Confidence Risks**:
- **[Risk]**: [percentage]% within [timeframe]
  - Prevention: [action items]

---

## DETAILED RESULTS

### Pattern Analysis
[Summary from meta_analyze]

**Top Patterns**:
1. [Pattern]: [correlation]% correlation
2. [Pattern]: [correlation]% correlation

### Suggestions Generated
[Summary from meta_learn]

**Quick Wins** ([N]):
- [Suggestion]: [ROI estimate]

**Implemented This Week** ([N]):
- [Suggestion]: [status]

### Documentation Updates
[Summary from meta_document]

**New Patterns Documented**: [N]
**Patterns Validated**: [N]/[N] passing

### Predictions
[Summary from meta_predict]

**High-Confidence**: [N] predictions
**Preventive Actions**: [N] recommended

### Experiments
[Summary from meta_experiment]

**Active**: [N]
**Promoted**: [N]
**Rolled Back**: [N]

### Agent Evolution
[Summary from meta_evolve]

**Agents Improved**: [N]
**Avg Improvement**: +[percentage]%
**Top Performer**: [agent-name] (+[percentage]%)

---

## RECOMMENDATIONS

### Immediate Action Required
1. [Action item with deadline]
2. [Action item with deadline]

### This Week's Focus
1. [Priority item]
2. [Priority item]

### Long-Term Initiatives
1. [Strategic initiative]
2. [Strategic initiative]

---

## NEXT WEEK

**Scheduled**:
- Next pipeline run: [date]
- Experiments to check: [list]
- Predictions to validate: [list]

**Manual Review Needed**:
- [Item requiring human decision]
- [Item requiring human decision]

---

**Report Generated**: [timestamp]
**Full Logs**: meta/logs/$SESSION_ID.log

**Commands for Details**:
- Analysis: `cat meta/weekly-analysis-$SESSION_ID.md`
- Suggestions: `cat meta/suggestions-$SESSION_ID.md`
- Health: `cat meta/health-$SESSION_ID.md`
- Evolution: `cat meta/evolution-$SESSION_ID.md`
- Predictions: `cat meta/predictions-$SESSION_ID.md`

Phase 10: Commit and Notify

echo ""
echo "Committing weekly improvements..."

if [ "$DRY_RUN" = false ]; then
  # Add all meta files
  git add meta/

  # Commit with comprehensive message
  COMMIT_MSG="meta: Weekly improvement pipeline - $(date +%Y-%m-%d)

Session: $SESSION_ID

Summary:
- Analyzed: $ANALYZE_STATUS
- Learned: $LEARN_STATUS
- Documented: $DOCUMENT_STATUS
- Predicted: $PREDICT_STATUS
- Experimented: $EXPERIMENT_STATUS
- Implemented: $IMPLEMENT_STATUS
- Evolved: $EVOLVE_STATUS
- Health: $HEALTH_STATUS

Details in meta/weekly-report-$SESSION_ID.md

Auto-generated by /meta_improve"

  git commit -m "$COMMIT_MSG"

  echo "✅ Changes committed"
  echo ""
  echo "To push: git push origin $(git branch --show-current)"
fi

echo ""
echo "=========================================="
echo "WEEKLY PIPELINE COMPLETE"
echo "=========================================="
echo ""
echo "Report: meta/weekly-report-$SESSION_ID.md"
echo "Logs: meta/logs/$SESSION_ID.log"
echo ""
echo "Summary:"
echo "  • Time saved this week: [X] hours"
echo "  • Improvements implemented: [N]"
echo "  • System velocity: [X]x"
echo "  • Health score: [N]/100"
echo ""
echo "Next run: [next scheduled date]"

Pipeline Configuration

Scheduling (Cron)

# Add to crontab for weekly Sunday 2am runs
0 2 * * 0 /path/to/claude /meta_improve >> meta/logs/cron.log 2>&1

Error Handling

Graceful Degradation:

  • If one phase fails, continue to next
  • Log all errors
  • Report failures in weekly summary
  • Never leave system in broken state

Rollback on Critical Failure:

  • If implement phase fails, rollback changes
  • If evolve phase corrupts agents, restore backups
  • All changes go through git for easy revert

Important Notes

  1. Run Weekly: Sunday 2am optimal (low traffic)
  2. Review Reports: Check weekly report every Monday
  3. Act on Predictions: Address high-confidence risks
  4. Monitor Health: Track compound ROI trends
  5. Iterate: System gets better each week
  6. Human Oversight: Review before merging improvements
  7. Celebrate Wins: Share successes with team

Example Usage Scenarios

Scenario 1: Full Weekly Run

/meta_improve
# Runs all 9 phases
# Generates comprehensive weekly report

Scenario 2: Test Pipeline (Dry-Run)

/meta_improve --dry-run
# Simulates pipeline without making changes
# Useful for testing

Scenario 3: Skip Expensive Phases

/meta_improve --skip evolve --skip experiment
# Runs faster pipeline
# Useful for mid-week check-ins

Scenario 4: Run Single Phase

/meta_improve --only analyze
# Runs only pattern analysis
# Useful for debugging specific phase

Remember: This pipeline embodies compound engineering - every week the system gets permanently better. Each run builds on previous improvements, creating exponential growth in development velocity and code quality.

Reference in New Issue View Git Blame Copy Permalink