Dead Code Detector

The Problem

Your Python project has 47 unused imports scattered across 23 files. Three functions haven't been called since the last refactor. There's a console.log('here') statement that made it through three code reviews. None of it breaks anything, all of it adds noise, and none of it will ever get cleaned up unless something forces it — because "it works" is the most dangerous phrase in software maintenance.

What You Get

• Multi-language detection — scans Python, JavaScript/TypeScript, Go, Rust, Java, and C/C++ with language-specific analysis (AST parsing for Python, type analysis for TS, compiler lint output for Rust)
• Unused import identification — finds imports that are never referenced, excluding type-only imports, side-effect imports, and dynamic usage patterns
• Dead function/class detection — locates functions and classes with zero references project-wide, accounting for framework conventions (decorators, route definitions, test discovery, dependency injection)
• Unreachable code detection — flags code after return, raise, break, continue, sys.exit(), and panic statements
• Debugging statement scanner — identifies console.log, debugger, print statements, and commented-out code blocks
• Confidence-scored report — categorizes findings as HIGH (provably unused), MEDIUM (unused in source but referenced in tests/docs), or LOW (dynamic usage possible) with removal recommendations

Why This Beats Prompting It Yourself

You can ask "find unused code" for a single file. But dead code accumulates across the entire codebase, in patterns that require project-wide cross-referencing — a function that looks unused might be called by a test, a decorator, or a dynamic dispatch. This skill performs the full cross-reference analysis with framework awareness, producing a report where every finding includes a confidence rating and specific removal guidance. Manual grep can't distinguish "unused" from "used via reflection."

Example Session

You: "Scan my Python project for dead code"

Agent: "Dead Code Report — src/ (142 files scanned)

Critical — Safe to Remove (8):

File                      Line   Type             Symbol                    Confidence
────────────────────────  ─────  ───────────────  ────────────────────────  ──────────
src/utils.py              12     Unused import    from collections import   HIGH
                                                  OrderedDict
src/api/handlers.py       45     Unused function  legacy_endpoint()         HIGH
src/models/old_schema.py  1-180  Dead module      entire file not imported  HIGH

Warning — Verify Before Removal (3):

File          Line  Type             Symbol            Confidence
────────────  ────  ───────────────  ────────────────  ─────────────────────────────────
src/auth.py   78    Unused function  oauth_callback()  MEDIUM (used via route decorator)
src/utils.py  92    Unused variable  temp_config       MEDIUM (referenced in tests only)

Info — Code Smells (5):

File          Line  Type             Symbol          Confidence
────────────  ────  ───────────────  ──────────────  ──────────
src/debug.py  5     Debug statement  print('debug')  HIGH

Summary: 16 dead code items found. 8 safe to remove immediately. 3 require manual verification. 5 code smells to clean up."

Use Cases

• Pre-release cleanup to reduce codebase bloat and cognitive load
• Code review gate to flag dead code introduced in PRs
• CI/CD quality gate to prevent dead code accumulation over time
• Post-migration cleanup after large feature removals or rewrites
• Onboarding audit to understand which parts of a legacy codebase are actually used

Known Limitations

Dynamic usage patterns (reflection, eval, getattr, dependency injection) produce false positives that are flagged as LOW confidence but may require manual verification. Framework-specific patterns like Django @app.route or React component registration need the full codebase context to avoid false removals. This skill does not remove code — it produces a report with removal recommendations.