Diff & Git No native equivalent

review_diff

Run 9 parallel static analysis checks on a git diff: secrets, breaking changes, coupling, complexity, dead-code, blast-radius, bug-patterns, test-gaps, hotspots.

What it does

review_diff runs nine independent static analysis checks against a git diff in parallel and returns a scored verdict: pass, warn, or fail.

The 9 checks

  1. Secrets — scans diff for leaked keys, tokens, credentials
  2. Breaking changes — detects removed or renamed public API surfaces
  3. Coupling gaps — finds changes that touch tightly coupled files without updating both
  4. Complexity — flags functions whose complexity increased
  5. Dead code — detects new exports with zero references
  6. Blast radius — measures how many downstream symbols are affected
  7. Bug patterns — finds common anti-patterns introduced in the diff
  8. Test gaps — flags changed production code without corresponding test changes
  9. Hotspots — warns when changes touch high-churn, high-complexity files

Output

A scored verdict with per-check results. Each check returns pass/warn/fail with evidence. The overall verdict is the worst individual result.

When to use it

  • Pre-push review — catch issues before they reach CI
  • PR review automation — structured analysis for code reviewers
  • Post-commit audit — verify recent changes didn’t introduce problems

Supports WORKING and STAGED as special refs for reviewing uncommitted changes.

Related tools

diff_outline changed_symbols impact_analysis

Benchmark note

This benchmark compares CodeSift against the closest practical native workflow an agent would use for the same task. For some tools, that baseline is a direct shell equivalent such as rg or find. For AST-aware, graph-aware, and LSP-backed tools, the baseline is a multi-step workflow rather than a strictly identical command. Results should be read as agent-workflow comparisons: token cost, call count, and practical context efficiency.