code-reduction

Overview

Systematic code reduction: discovers redundancy, removes dead code, identifies duplicates, finds gaps, and consolidates. Two modes available:

• Quick scan (default for targeted cleanup) — runs knip, depcheck, or vulture for fast tool-driven dead code detection
• Deep analysis (default for broad scope) — spawns parallel Explore agents for structural analysis, deduplication, and gap detection
• All — runs quick scan first, then escalates to deep analysis for anything the tools missed

Required References

Before starting, read references/methodology/refactor-planning.md for safe refactoring strategies and dependency analysis patterns.

Step 0: Quick Scan (Tool-Driven)

Detect project type and run the appropriate static analysis tools:

Signal	Tools available
package.json with TypeScript	knip, ts-prune, depcheck
package.json without TypeScript	depcheck
pyproject.toml	vulture, pip-extra-reqs
go.mod	go vet, staticcheck

# TypeScript/Node.js — knip is the best all-in-one tool
npx knip --reporter compact    # Finds unused files, exports, deps, types

# Fallbacks if knip is not available:
npx depcheck                   # Unused deps only

# Python
pip install vulture && vulture src/

# Go
go vet ./...
staticcheck ./...

Triage quick scan results

Category	Action
Truly unused — no references anywhere	Delete
Dynamically used — referenced via string, reflection, or config	Keep, add to knip ignore
Test-only export — only used in tests	Keep (but check if test is dead too)
Plugin/framework magic — used by framework convention	Keep, add to knip ignore

If mode is quick, apply deletions (Step 4) and report (Step 5). If mode is deep or all, continue to Step 1.

Step 1: Discovery (Parallel — 3 Agents)

Spawn 3 Explore agents to scan in parallel:

1. Agent A: Map exports, imports, and dependency graph
2. Agent B: Identify code patterns and potential duplicates
3. Agent C: Catalog test coverage and unused test helpers