
cleanup-health-inline
by maslennikov-ig
๐ผ Turn Claude Code into a production powerhouse. 33+ AI agents automate bug fixing, security scanning, and dependency management. 19 slash commands, 6 MCP configs (600-5000 tokens), quality gates, and health monitoring. Ship faster, ship safer, ship smarter.
SKILL.md
name: cleanup-health-inline description: Inline orchestration workflow for dead code detection and removal. Provides step-by-step phases for dead-code-hunter detection, priority-based cleanup with dead-code-remover, and verification cycles. version: 2.0.0
Cleanup Health Check (Inline Orchestration)
You ARE the orchestrator. Execute this workflow directly without spawning a separate orchestrator agent.
Workflow Overview
Detection โ Validate โ Remove by Priority โ Verify โ Repeat if needed
Max iterations: 3 Priorities: critical โ high โ medium โ low
Phase 1: Pre-flight
-
Setup directories:
mkdir -p .tmp/current/{plans,changes,backups} -
Validate environment:
- Check
package.jsonexists - Check
type-checkandbuildscripts exist
- Check
-
Initialize TodoWrite:
[ {"content": "Dead code detection", "status": "in_progress", "activeForm": "Detecting dead code"}, {"content": "Remove critical dead code", "status": "pending", "activeForm": "Removing critical dead code"}, {"content": "Remove high priority dead code", "status": "pending", "activeForm": "Removing high dead code"}, {"content": "Remove medium priority dead code", "status": "pending", "activeForm": "Removing medium dead code"}, {"content": "Remove low priority dead code", "status": "pending", "activeForm": "Removing low dead code"}, {"content": "Verification scan", "status": "pending", "activeForm": "Verifying cleanup"} ]
Phase 2: Detection
Invoke dead-code-hunter via Task tool:
subagent_type: "dead-code-hunter"
description: "Detect all dead code"
prompt: |
Scan the entire codebase for dead code:
- Unused imports and exports
- Commented out code blocks
- Unreachable code
- Debug statements (console.log, debugger)
- Unused variables and functions
- Unused dependencies
- Categorize by priority (critical/high/medium/low)
Generate: dead-code-report.md
Return summary with dead code counts per priority.
After dead-code-hunter returns:
- Read
dead-code-report.md - Parse dead code counts by priority
- If zero dead code โ skip to Final Summary
- Update TodoWrite: mark detection complete
Phase 3: Quality Gate (Detection)
Run inline validation:
pnpm type-check
pnpm build
- If both pass โ proceed to removal
- If fail โ report to user, exit
Phase 4: Removal Loop
For each priority (critical โ high โ medium โ low):
-
Check if dead code exists for this priority
- If zero โ skip to next priority
-
Update TodoWrite: mark current priority in_progress
-
Invoke dead-code-remover via Task tool:
subagent_type: "dead-code-remover" description: "Remove {priority} dead code" prompt: | Read dead-code-report.md and remove all {priority} priority dead code. For each item: 1. Backup file before editing 2. Remove dead code 3. Log change to .tmp/current/changes/cleanup-changes.json Generate/update: dead-code-cleanup-summary.md Return: count of removed items, count of failed removals. -
Quality Gate (inline):
pnpm type-check pnpm build- If FAIL โ report error, suggest rollback, exit
- If PASS โ continue
-
Update TodoWrite: mark priority complete
-
Repeat for next priority
Phase 5: Verification
After all priorities cleaned:
-
Update TodoWrite: mark verification in_progress
-
Invoke dead-code-hunter (verification mode):
subagent_type: "dead-code-hunter" description: "Verification scan" prompt: | Re-scan codebase after cleanup. Compare with previous dead-code-report.md. Report: - Dead code removed (count) - Dead code remaining (count) - New dead code introduced (count) -
Decision:
- If dead_code_remaining == 0 โ Final Summary
- If iteration < 3 AND dead_code_remaining > 0 โ Go to Phase 2
- If iteration >= 3 โ Final Summary with remaining items
Phase 6: Final Summary
Generate summary for user:
## Cleanup Health Check Complete
**Iterations**: {count}/3
**Status**: {SUCCESS/PARTIAL}
### Results
- Found: {total} dead code items
- Removed: {removed} ({percentage}%)
- Remaining: {remaining}
### By Priority
- Critical: {removed}/{total}
- High: {removed}/{total}
- Medium: {removed}/{total}
- Low: {removed}/{total}
### Validation
- Type Check: {status}
- Build: {status}
### Artifacts
- Detection: `dead-code-report.md`
- Cleanup: `dead-code-cleanup-summary.md`
Error Handling
If quality gate fails:
Rollback available: .tmp/current/changes/cleanup-changes.json
To rollback:
1. Read changes log
2. Restore files from .tmp/current/backups/
3. Re-run workflow
If worker fails:
- Report error to user
- Suggest manual intervention
- Exit workflow
Key Differences from Old Approach
| Old (Orchestrator Agent) | New (Inline Skill) |
|---|---|
| 9+ orchestrator calls | 0 orchestrator calls |
| ~1400 lines (cmd + agent) | ~150 lines |
| Context reload each call | Single session context |
| Plan files for each phase | Direct execution |
| ~10,000+ tokens overhead | ~500 tokens |
Worker Prompts
See references/worker-prompts.md for detailed prompts.
Score
Total Score
Based on repository quality metrics
SKILL.mdใใกใคใซใๅซใพใใฆใใ
ใฉใคใปใณในใ่จญๅฎใใใฆใใ
100ๆๅญไปฅไธใฎ่ชฌๆใใใ
GitHub Stars 100ไปฅไธ
3ใถๆไปฅๅ ใซๆดๆฐใใใ
10ๅไปฅไธใใฉใผใฏใใใฆใใ
ใชใผใใณIssueใ50ๆชๆบ
ใใญใฐใฉใใณใฐ่จ่ชใ่จญๅฎใใใฆใใ
1ใคไปฅไธใฎใฟใฐใ่จญๅฎใใใฆใใ
Reviews
Reviews coming soon