Code Analyzer Skill
Installation
No separate download: the skill runs the in-repo tool .claude/tools/analysis/project-analyzer/analyzer.mjs.
- Ensure Node.js (v18+) is installed: nodejs.org or
winget install OpenJS.NodeJS.LTS(Windows),brew install node(macOS). - From the project root, the script is invoked automatically; no extra install steps.
Cheat Sheet & Best Practices
Metrics: Focus on cyclomatic complexity (decision paths), LOC, maintainability index, and duplicate blocks. Use ESLint complexity rule (e.g. "complexity": ["error", 15]) for JS/TS; optional chaining and default params add branches.
Process: Analyze before refactoring; run project-wide then drill into hotspots. Track trends over time (not one-off). Use max-depth, max-lines, max-nested-callbacks, max-params, max-statements alongside complexity.
Hacks: Start with project-analyzer output; filter by file type and threshold. Prioritize files with high complexity and high churn. Disable complexity rule only if you cannot set a sensible limit; prefer lowering the threshold over disabling.
Certifications & Training
No single cert; aligns with static analysis and ESLint complexity. ESLint: complexity rule, max-depth, max-lines, max-params. Skill data: Cyclomatic complexity, LOC, maintainability, duplicates; analyze before refactor; track hotspots and trends.
Hooks & Workflows
Suggested hooks: Pre-commit or CI: run project-analyzer/doctor for health; optional complexity gate. Use with developer (secondary), qa (secondary), code-reviewer (primary).
Workflows: Use with code-reviewer (primary), developer/ qa (secondary), c4-code (primary). Flow: run analyzer → filter hotspots → refactor or add tests. See code-review-workflow.md.
Overview
Static code analysis and metrics. 90%+ context savings.
Tools (Progressive Disclosure)
Analysis
| Tool | Description |
|---|---|
| analyze-file | Analyze single file |
| analyze-project | Analyze entire project |
| complexity | Calculate complexity metrics |
Metrics
| Tool | Description |
|---|---|
| loc | Lines of code |
| cyclomatic | Cyclomatic complexity |
| maintainability | Maintainability index |
| duplicates | Find duplicate code |
Reporting
| Tool | Description |
|---|---|
| summary | Get analysis summary |
| hotspots | Find complexity hotspots |
| trends | Analyze metric trends |
Agent Integration
- code-reviewer (primary): Code review
- refactoring-specialist (primary): Tech debt analysis
- architect (secondary): Architecture assessment
Iron Laws
- ALWAYS run project-wide analysis before drilling into individual files — local analysis without context misses which files are actually the highest-priority hotspots; start broad, then focus.
- ALWAYS focus on high-complexity AND high-churn files — a complex but rarely-changed file is lower priority than a moderately complex but frequently-changed one; intersection matters most.
- NEVER set complexity thresholds above 20 — cyclomatic complexity >20 is demonstrably correlated with defects; teams that allow >20 accumulate unmaintainable code without noticing.
- ALWAYS track metrics over time, not just once — a single analysis snapshot is meaningless; track trends weekly to detect gradual degradation before it becomes a crisis.
- NEVER report metrics without actionable next steps — complexity numbers without refactoring targets provide no value; every high-complexity finding must include a specific suggested improvement.
Anti-Patterns
| Anti-Pattern | Why It Fails | Correct Approach |
|---|---|---|
| Analyzing only changed files | Misses cross-file complexity accumulation | Run project-wide then filter to changed hot spots |
| Ignoring high-complexity files over time | Gradual degradation invisible in point-in-time analysis | Track weekly trends; alert on any increase |
| Complexity threshold >20 | Research shows defect rate spikes sharply above 20 | Set ESLint complexity rule to ≤15 for enforcement |
| Reporting metrics without action items | Metrics without remediation don't reduce complexity | Attach specific refactoring suggestion per hotspot |
| Running analysis once and ignoring results | Technical debt silently accumulates | Schedule automated weekly analysis with trend reports |
Memory Protocol (MANDATORY)
Before starting:
Read .claude/context/memory/learnings.md
After completing:
- New pattern ->
.claude/context/memory/learnings.md - Issue found ->
.claude/context/memory/issues.md - Decision made ->
.claude/context/memory/decisions.md
ASSUME INTERRUPTION: If it's not in memory, it didn't happen.