Dead Code Review

Dead Code Review (Copilot Edition — Sequential)

Six review phases done in order. Unlike the Claude canonical (which spawns parallel workers, Sonnet-first and tiered by load), this version walks them yourself, one at a time. Slower but complete. The model-tier cap (skills/sdlc/templates/models.md) is advisory here — phases run inline in your session model, so set your session model for the savings.

Scope flag

Phase 0 — Test baseline

Run the configured test suites (from .claude/project.json test.* keys) and record pass/fail counts. This is the baseline — any removal must keep these counts intact.

Phase 1 — Backend

Scan backend files for:

• Unused imports (grep each import against file usage).
• Dead functions (defined but not called anywhere in the repo).
• Dead API endpoints (routes with no frontend or external consumer).
• Unused schema classes (models / DTOs / Pydantic models never referenced).
• Stale one-time scripts (run_*.py, seed_*.py, diagnostic scripts already run).
• Commented-out code blocks older than current work.

Phase 2 — Frontend

Scan frontend source for:

• Unused components (grep import name across all files).
• Unused hooks / lib exports / type definitions.
• Dead store properties or actions.
• Dead API methods or types in the API client module.
• Unused dependencies in package.json.
• Stale test files for removed components.

Phase 3 — Database & migrations

Scan migration files and query live schema if available:

• Unused tables (exist in DB but never referenced in code).
• Unused columns (never SELECTed / INSERTed / UPDATEd).
• Duplicate indexes (regular index on same columns as a UNIQUE constraint).
• Redundant migrations (CREATE then DROP pairs; ALTER adding columns that already exist).
• Empty tables that may indicate abandoned features.

Phase 4 — Docs & plans

Scan markdown files in docs/, plans/, and root:

• Completed plans (feature fully shipped → delete).
• Stale root markdown (one-time setup guides, old debugging notes).
• Outdated docs that conflict with current CLAUDE.md / AGENTS.md.
• Empty directories left from prior cleanup.

Phase 5 — Scripts & config

• One-time scripts already run (data seeders, migration fixers, diagnostic checks).
• Stale test files for removed features.
• Orphaned config files (unused build configs, old auth state).
• Generated output directories that should be gitignored.

Triage & execute

For each finding, assign confidence:

• HIGH — demonstrably unused (grep confirms zero references). Remove.
• MEDIUM — likely unused but harder to prove (string-dispatch, reflection). Leave for user approval.
• LOW — flagged as "looks suspicious" but uncertain. Report only; don't remove.

Execute HIGH-confidence removals first. Restart services if backend changed. Re-run the test suite — if green, execute MEDIUM items that the user explicitly approves. Final test run to confirm.

Report

• Files deleted (count + total lines).
• Files modified (count + lines removed).
• Database objects dropped (tables / columns / indexes).
• Test comparison: baseline vs post-cleanup.

Rules

• Never commit during the review — the user decides when.
• Always baseline and re-run tests around the changes.
• Only remove at HIGH confidence unless the user explicitly approves MEDIUM items.
• Check GOTCHAS.md Code Hygiene section for known-safe patterns.
• Do not remove functions that look unused but are referenced by string-based dispatch (intent routing, plugin registries).
• Do not remove a type that is the return type of a live API method, even if never imported externally.