| name | audit-refactoring |
| description | Run a single-session refactoring audit on the codebase |
| supports_parallel | true |
| fallback_available | true |
| estimated_time_parallel | 20 min |
| estimated_time_sequential | 60 min |
Single-Session Refactoring Audit
When to Use
- Tasks related to audit-refactoring
- User explicitly invokes
/audit-refactoring
When NOT to Use
- When the task doesn't match this skill's scope -- check related skills
- When a more specialized skill exists for the specific task
Purpose
This audit identifies technical debt related to code structure, duplication, and
complexity that requires refactoring. It focuses on five key areas:
- God Objects - Large files with too many responsibilities
- Code Duplication - Repeated patterns and copy-paste code
- Cognitive Complexity - Functions/methods exceeding complexity thresholds
- Architecture Violations - Layer boundary violations and import cycles
- Technical Debt Markers - TODOs, FIXMEs, and HACKs requiring cleanup
When to use this audit:
- Triggered by refactoring-specific thresholds (40+ commits, 3+ complexity
warnings, circular dependencies detected)
- Before major refactoring initiatives to establish baselines
- After rapid feature development to identify accumulated structural debt
Scope:
- Include:
src/, components/, lib/, hooks/
- Exclude:
node_modules/, dist/, build/, docs/
Execution Mode Selection
| Condition | Mode | Time |
|---|
| Task tool available + no context pressure | Parallel | ~20 min |
| Task tool unavailable | Sequential | ~60 min |
| Context running low (<20% remaining) | Sequential | ~60 min |
| User requests sequential | Sequential | ~60 min |
Section A: Parallel Architecture (3 Agents)
When to use: Task tool available, sufficient context budget
Agent 1: structure-and-duplication
Focus Areas:
- God Objects (files > 300 lines, excessive responsibilities)
- Code Duplication (repeated patterns, copy-paste code)
- Large functions (> 50 lines, complexity risk)
Files:
src/**/*.tsx, components/**/*.tsx
lib/**/*.ts, hooks/**/*.ts
- Files flagged by
wc -l analysis
Targets:
- Files > 300 lines (potential split candidates)
- Functions > 50 lines (complexity risk)
- Components with > 10 props (interface too large)
- Duplicate code patterns (>10 lines repeated)
Agent 2: complexity-and-coupling
Focus Areas:
- Cognitive Complexity (CRITICAL targets)
- Architecture Violations (layer boundaries, import cycles)
- High coupling (files with > 5 imports from different domains)
Files:
- Files with high complexity (from static analysis tools)
- Files in circular dependency chains
- Cross-domain coupling patterns
Targets:
- CRITICAL cognitive complexity issues
- Circular dependencies
- Cross-layer imports violating architecture boundaries
- High fan-in/fan-out modules
Agent 3: dead-code-and-markers
Focus Areas:
- Technical Debt Markers (TODOs, FIXMEs, HACKs)
- Unused exports
- Dead code patterns
Files:
- All source files with TODO/FIXME/HACK comments
- Commented-out code blocks
Targets:
- TODO/FIXME/HACK markers requiring action
- Commented-out code blocks
- Orphaned test files or utilities
Parallel Execution Command
Invoke all 3 agents in a SINGLE Task message:
Task 1: structure-and-duplication agent - audit god objects and code duplication
Task 2: complexity-and-coupling agent - audit cognitive complexity and
architecture violations Task 3: dead-code-and-markers agent - audit tech debt
markers and dead code
Each agent prompt MUST end with:
CRITICAL RETURN PROTOCOL:
- Write findings to the specified output file using Write tool or Bash
- Return ONLY: `COMPLETE: [agent-id] wrote N findings to [output-path]`
- Do NOT return full findings content — orchestrator checks completion via file
Dependency constraints: All 3 agents are independent -- no ordering
required. Each writes to a separate JSONL section. Results are merged after all
agents complete.
Coordination Rules
- Each agent writes findings to separate JSONL section
- Dead code findings have lowest priority in conflicts
- Complexity findings have highest priority
- Architecture agent handles boundary overlap issues
Section B: Sequential Fallback (Single Agent)
When to use: Task tool unavailable, context limits, or user preference
Execution Order:
- Cognitive Complexity Analysis - 15 min
- God Objects & Duplication - 15 min
- Architecture Violations - 15 min
- Tech Debt Markers & Dead Code - 15 min
Total: ~60 min (vs ~20 min parallel)
Checkpoint Format
{
"started_at": "ISO timestamp",
"categories_completed": ["Complexity", "GodObjects"],
"current_category": "Architecture",
"findings_count": 12,
"last_file_written": "stage-2-findings.jsonl"
}
Pre-Audit Validation
Step 0: Episodic Memory Search
Before running refactoring audit, search for context from past sessions:
mcp__plugin_episodic_memory_episodic_memory__search({
query: ['refactoring audit', 'god object', 'complexity'],
limit: 5,
});
mcp__plugin_episodic_memory_episodic_memory__search({
query: ['cognitive complexity', 'duplicate code', 'circular'],
limit: 5,
});
Why this matters:
- Compare against previous refactoring targets
- Identify recurring complexity hotspots (may need architectural fix)
- Track which files were flagged before and why
- Prevent re-flagging intentional design decisions
Step 1: Check Thresholds
Run npm run review:check and report results.
- If no thresholds triggered: "No review thresholds triggered. Proceed anyway?"
- Continue with audit regardless (user invoked intentionally)
Step 2: Gather Current Baselines
Collect these metrics by running commands:
npm run deps:circular 2>&1 || echo "No circular dep tool configured"
find src components lib -name "*.ts" -o -name "*.tsx" 2>/dev/null | xargs wc -l 2>/dev/null | sort -n | tail -20
grep -rn "TODO\|FIXME\|HACK" --include="*.ts" --include="*.tsx" 2>/dev/null | wc -l
Step 3: Load False Positives Database
Read docs/technical-debt/FALSE_POSITIVES.jsonl and filter findings matching:
- Category:
refactoring
- Expired entries (skip if
expires date passed)
Note patterns to exclude from final findings.
Step 4: Check Prior Audit Results
Check docs/audits/single-session/refactoring/ for previous audit results:
- Compare baselines (file counts, complexity scores, circular deps)
- Identify regressed issues (previously fixed, now reappeared)
- Note resolved findings to avoid duplicate work
Audit Execution
Focus Areas (5 Categories):
- God Objects - Large files with too many responsibilities
- Files > 300 lines (potential split candidates)
- Classes/modules with > 5 distinct responsibilities
- Components with > 10 props (interface too large)
- Code Duplication - Repeated patterns and copy-paste code
- Multiple similar code blocks (>10 lines duplicated)
- Repeated business logic across components
- Pattern opportunities for extraction
- Cognitive Complexity - CRITICAL targets
- Functions with high cyclomatic complexity
- Deep nesting (>3 levels)
- Functions > 50 lines
- Architecture Violations - Layer boundaries and import cycles
- Circular dependencies
- Files with > 5 imports from different domains (coupling)
- Cross-layer imports violating architecture boundaries
- Technical Debt Markers - TODOs, FIXMEs, HACKs
- TODO/FIXME/HACK markers requiring action
- Commented-out code blocks
For each category:
- Search relevant files using Grep/Glob
- Identify specific issues with file:line references
- Classify severity: S0 (blocking) | S1 (major friction) | S2 (annoying) | S3
(nice-to-have)
- Estimate effort: E0 (trivial) | E1 (hours) | E2 (day) | E3 (major)
- Assign confidence level (see Evidence Requirements below)
Output Format
Category Field: All findings MUST use category: "refactoring"
Output Directory: docs/audits/single-session/refactoring/audit-YYYY-MM-DD/
1. Markdown Summary (display to user):
## Refactoring Audit - [DATE]
### Baselines
- Circular dependencies: X
- Files > 300 lines: X
- TODO/FIXME/HACK markers: X
### Findings Summary
| Severity | Count | Category | Confidence |
| -------- | ----- | -------- | ----------- |
| S0 | X | ... | HIGH/MEDIUM |
| S1 | X | ... | HIGH/MEDIUM |
| S2 | X | ... | ... |
| S3 | X | ... | ... |
### Top Refactoring Candidates
1. [file] - X lines, Y responsibilities (S1/E2) - DUAL_PASS_CONFIRMED
2. ...
### False Positives Filtered
- X findings excluded (matched FALSE_POSITIVES.jsonl patterns)
### Quick Wins (E0-E1)
- ...
### Batch Fix Opportunities
- X instances of [pattern] can be auto-fixed
- ...
### Recommendations
- ...
2. JSONL Findings (save to file):
Create file:
docs/audits/single-session/refactoring/audit-[YYYY-MM-DD]/findings.jsonl
Category field: category MUST be refactoring. Also include symbols and
optionally duplication_cluster for duplication findings.
3. Markdown Report (save to file):
Create file:
docs/audits/single-session/refactoring/audit-[YYYY-MM-DD]/REPORT.md
Full markdown report with all findings, baselines, and refactoring plan.
Standard Audit Procedures
Read CANON/templates/AUDIT_TEMPLATE.md for: Evidence Requirements,
Dual-Pass Verification, Cross-Reference Validation, JSONL Output Format,
Context Recovery, Post-Audit Validation, MASTER_DEBT Cross-Reference,
Interactive Review, TDMS Intake & Commit, Documentation References, Agent
Return Protocol, and Honesty Guardrails.
Skill-specific TDMS intake:
node scripts/debt/intake-audit.js <output.jsonl> --source "audit-refactoring-<date>"
Refactoring audit triggers (check AUDIT_TRACKER.md):
- 40+ commits since last refactoring audit, OR
- 3+ new complexity warnings, OR
- Circular dependency detected
Version History
| Version | Date | Description |
|---|
| 1.0 | 2026-02-25 | Initial implementation |