Agent Skills: Find specs missing constitutional alignment

You are the housekeeping workflow coordinator. Your job is to ensure the project's documentation and issue tracking accurately reflects reality, organize completed work, and provide a clear snapshot of current state and next steps.

Core Responsibilities

Execution Strategy: Multi-Phase Workflow

This is a workflow skill - it coordinates multiple subagents in parallel for maximum efficiency.

Phase 1: Governance Refresh (Sequential)

Why sequential: Need current state understanding before dispatching parallel agents.

Task: Update governance documents to reflect current reality

Approach:

1.1 Refresh Governance Metrics

Use the refreshing-metrics skill (invoke with Skill tool) to update:

• .wrangler/issues/README.md - Issue counts and status percentages
• .wrangler/specifications/README.md - Spec counts and constitutional compliance
• .wrangler/ROADMAP_NEXT_STEPS.md - Implementation status and overall % complete

1.2 Update Roadmap Next Steps

After metrics refresh, manually review and update Next Steps:

1. Read .wrangler/ROADMAP_NEXT_STEPS.md
2. Review recent git commits (last 20-30 commits) for completed work
3. Review recently closed issues and specifications
4. Update three categories:
   • Move ⚠️ → ✅ if features fully completed
   • Move ❌ → ⚠️ if features started implementation
   • Add new ❌ items if gaps identified
5. Update "What Works Well" and "Critical Gap" sections
6. Update "Prioritized Roadmap" if priorities shifted
7. Update "Last Updated By" and "Next Review" dates

1.3 Verify Constitutional Compliance

Check if any specifications lack constitutional alignment:

# Find specs missing constitutional alignment
grep -L "Constitutional Alignment" .wrangler/specifications/*.md | grep -v "_CONSTITUTION\|_ROADMAP\|README"

If found, note for Agent C (Documentation Drift) to flag.

Output:

• Updated governance metrics
• Updated roadmap Next Steps
• List of specs needing constitutional alignment sections

Phase 2: Parallel Issue Reconciliation (Parallel)

Why parallel: Five independent tasks with no dependencies - run simultaneously for speed.

Launch five parallel subagents using the Task tool:

Agent A: Open Issues Reconciliation

Task: Review all open issues and ensure they match implementation reality

Approach:

1. List all open issues (issues_list with status: ["open", "in_progress"])
2. For each open issue:
   • Read the issue content
   • Check if described work has actually been completed
   • Check if requirements have changed during implementation
   • Identify issues that should be closed but weren't
   • Identify issues that need updates due to design changes
3. Take action:
   • If completed but not closed: Use issues_mark_complete or issues_update to close
   • If requirements drifted: Use issues_update to reflect actual implementation
   • If blocked/cancelled: Use issues_update to mark status appropriately
4. Track metrics:
   • Issues reviewed: [count]
   • Issues closed: [count]
   • Issues updated: [count]
   • Issues with drift: [count]

Output: List of actions taken and current state summary

Agent B: Completed Issues Organization

Task: Move completed/closed issues to .wrangler/issues/archived/ directory for archival

Approach:

1. List all closed issues (issues_list with status: ["closed", "cancelled"])
2. Check if .wrangler/issues/archived/ directory exists, create if not
3. For each closed issue:
   • Check if still in .wrangler/issues/ root directory
   • If yes, move to .wrangler/issues/archived/
   • Preserve filename (don't rename)
4. Track metrics:
   • Completed issues found: [count]
   • Issues moved: [count]
   • Already organized: [count]

Output: Organization summary with file counts

Agent C: Documentation Drift Detection

Task: Identify areas where documentation may not reflect implementation reality, including governance compliance

Approach:

1. Read key documentation files:
   • README.md
   • CLAUDE.md (or equivalent project documentation)
   • .wrangler/CONSTITUTION.md
   • .wrangler/ROADMAP.md
   • Any specification files
2. Review recent git commits for major changes
3. Look for signs of drift:
   • Features mentioned in docs but not implemented
   • Features implemented but not documented
   • API changes not reflected in docs
   • Configuration changes not updated
   • File structure changes not documented
   • Specifications missing constitutional alignment sections
   • Roadmap phases not reflected in Next Steps
   • Constitutional principles violated in recent code
4. Create list of documentation updates needed
5. Track metrics:
   • Documentation files reviewed: [count]
   • Drift issues found: [count]
   • Constitutional compliance gaps: [count]
   • Severity: High/Medium/Low

Output: List of documentation drift issues with recommendations, including governance compliance issues

Agent D: Root Directory Organization

Task: Clean up and organize markdown files at project root

Approach:

1. Scan project root for organizational candidates:
   • RCA-.md, ANALYSIS-.md, IMPLEMENTATION-*.md, etc.
   • Any ALL_CAPS.md files (excluding standard files like README.md)
2. Categorize each file:
   • Delete if obsolete (discarded analysis, temp notes)
   • Move to memos/ if reference material (RCA, lessons learned, summaries)
   • Move to docs/ if user-facing documentation
   • Move to devops/docs/ if developer/maintainer documentation
3. Execute actions (delete or move with appropriate naming)
4. Track metrics:
   • Files processed: [count]
   • Deleted: [count]
   • Moved to memos/: [count]
   • Moved to docs/: [count]
   • Moved to devops/docs/: [count]

Output: Organized root directory with action summary

Agent E: Dangling Worktree Cleanup

Task: Remove git worktrees whose associated feature PRs have been merged

Approach:

1. List all git worktrees (git worktree list --porcelain)
2. For each feature worktree (skip main):
   • Check for uncommitted changes (git status --porcelain)
   • Check for unpushed commits (git log @{upstream}..HEAD)
   • Find associated PR (gh pr list --head "$BRANCH" --state all)
   • ONLY if PR is MERGED: mark for cleanup
3. For worktrees marked for cleanup:
   • Remove worktree (git worktree remove /path)
   • Delete merged branch (git branch -d branch-name)
4. Prune stale worktree metadata (git worktree prune)
5. Track metrics:
   • Worktrees checked: [count]
   • Worktrees removed: [count]
   • Worktrees kept: [count] (with reasons)
   • Branches deleted: [count]

CRITICAL SAFETY RULES:

• NEVER remove a worktree if PR status cannot be confirmed
• NEVER remove if PR is open or closed-without-merge
• NEVER remove if worktree has uncommitted changes or unpushed commits
• NEVER remove if no PR exists (might be active development)
• Default to KEEP if any uncertainty

Output: Worktree cleanup summary with kept/removed counts and reasons

Phase 3: Summary Report (Sequential)

Why sequential: Needs results from all Phase 2 agents.

Task: Compile comprehensive housekeeping report

Approach:

1. Wait for all Phase 2 agents to complete
2. Aggregate metrics from all agents
3. Create summary report showing:
   • Roadmap status (updated/current)
   • Issues reconciled (count, actions taken)
   • Files organized (count, location)
   • Documentation drift (issues found, severity)
   • Time taken per phase (if trackable)
   • Recommendations for follow-up work

Output: Housekeeping summary report

Workflow Execution Plan

Step 1: Start Phase 1 (Governance Refresh)

Execute governance refresh yourself:

1. Invoke refreshing-metrics skill using Skill tool
2. Manually update Next Steps with recent completions
3. Check for specs missing constitutional alignment

Step 2: Launch Phase 2 Parallel Agents

Use five separate Task tool calls in a single message to dispatch parallel agents:

I'm launching five parallel housekeeping agents:

[Use Task tool - Agent A: Open Issues Reconciliation]
[Use Task tool - Agent B: Completed Issues Organization]
[Use Task tool - Agent C: Documentation Drift Detection]
[Use Task tool - Agent D: Root Directory Organization]
[Use Task tool - Agent E: Dangling Worktree Cleanup]

CRITICAL: All five Task tool calls must be in a single response to execute truly in parallel.

Step 3: Collect Results & Generate Report

Wait for all agents to complete, then compile summary report.

Housekeeping Report Template

# Housekeeping Report - [Date]

## Summary

Housekeeping workflow completed successfully.

**Duration:** [X] minutes total

- Phase 1 (Governance refresh): [X] minutes
- Phase 2 (Parallel reconciliation): [X] minutes
- Phase 3 (Report generation): [X] minutes

## Governance Refresh

✅ **Metrics Updated**:

- `.wrangler/issues/README.md` - Updated issue counts and status percentages
- `.wrangler/specifications/README.md` - Updated spec counts and constitutional compliance
- `.wrangler/ROADMAP_NEXT_STEPS.md` - Updated implementation status

**Overall Project Completion**: ~[X]% ([+/-]% since last housekeeping)

**Completed since last housekeeping:**

- [Item 1]
- [Item 2]

**Current priorities:**

- [Priority 1]
- [Priority 2]

**Blockers identified:**

- [Blocker 1 if any]

**Constitutional Compliance**:

- Specifications with alignment: [X]/[Y] ([Z]%)
- Specifications needing alignment sections: [count]

## Issue Reconciliation

**Agent A - Open Issues:**

- Issues reviewed: [count]
- Issues closed: [count]
- Issues updated for drift: [count]
- Status: ✅ Complete

**Agent B - Completed Issues Organization:**

- Completed issues found: [count]
- Issues moved to archive: [count]
- Status: ✅ Complete

**Agent C - Documentation Drift:**

- Documentation files reviewed: [count]
- Drift issues found: [count]
  - High severity: [count]
  - Medium severity: [count]
  - Low severity: [count]
- Constitutional compliance gaps: [count]
- Status: ✅ Complete

**Agent D - Root Directory Organization:**

- Files processed: [count]
- Deleted (obsolete): [count]
- Moved to memos/: [count]
- Moved to docs/: [count]
- Moved to devops/docs/: [count]
- Status: ✅ Complete

**Agent E - Dangling Worktree Cleanup:**

- Worktrees checked: [count]
- Worktrees removed (PR merged): [count]
- Worktrees kept: [count]
- Branches deleted: [count]
- Status: ✅ Complete

## Actions Taken

### Issues Closed

- #000XXX - [Issue title] - Reason: [completed/cancelled]

### Issues Updated

- #000XXX - [Issue title] - Change: [what was updated]

### Files Organized

- Moved [X] issues to `issues/archived/`

### Worktrees Cleaned Up

- Removed [X] worktrees (merged PRs)
- Kept [Y] worktrees (active development)

### Documentation Drift Issues

**High Priority:**

- [ ] [Issue 1 description]

**Medium Priority:**

- [ ] [Issue 2 description]

**Low Priority:**

- [ ] [Issue 3 description]

## Recommendations

1. [Recommendation 1]
2. [Recommendation 2]

## Metrics

- Total issues in system: [count]
- Open issues: [count]
- In progress: [count]
- Completed: [count]
- Issue completion rate: [X]%
- Documentation coverage: [assessment]

---

_Housekeeping workflow v1.0 - Generated by wrangler:housekeeping_

Usage Examples

Example 1: Regular Housekeeping

User: "Run housekeeping"

Agent Response:

1. "Starting housekeeping workflow..."
2. "Phase 1: Updating .wrangler/ROADMAP_NEXT_STEPS.md..."
3. "Phase 2: Launching five parallel reconciliation agents..."
   • Dispatches Agent A (issues)
   • Dispatches Agent B (organization)
   • Dispatches Agent C (drift detection)
   • Dispatches Agent D (root directory)
   • Dispatches Agent E (worktree cleanup)
4. Waits for completion
5. "Phase 3: Generating housekeeping report..."
6. Presents summary report

Example 2: Post-Sprint Housekeeping

User: "We just finished the authentication feature sprint, run housekeeping"

Agent Response:

1. Updates roadmap with authentication feature completion
2. Reconciles auth-related issues
3. Moves completed auth issues to archive
4. Checks if auth documentation is up to date
5. Reports on sprint completion status

Workflow Metrics & Observability

Track these metrics for workflow optimization:

Efficiency Metrics:

• Total housekeeping time
• Time per phase
• Parallel speedup factor (Phase 2 time vs. sequential equivalent)

Action Metrics:

References

For detailed information, see:

• references/detailed-guide.md - Complete workflow details, examples, and troubleshooting