Review gate

Skill cukas/claude-workflow-skills/skills/review-gate

Guarded development workflows for Claude Code — investigation before fixes, orchestration before implementation, verification before completion

Install

npx -y skills add cukas/claude-workflow-skills --skill review-gate

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Use after completing any code implementation, bug fix, or refactor. Structured review with pass/fail verification before declaring done.

SKILL.md

2.6 KB, 592 tokens by cl100k_base, as published. Nobody here has run it

Review-Gate

Structured post-implementation review. Nothing passes without evidence.

1. Trace affected code paths

Use Read, Grep, and Glob to trace the call chain from each changed function to its callers and dependents. For changes that touch multiple files or cross API boundaries, trace both directions.

2. Bug pattern scan

Check every changed file for:

Null safety: optional chaining gaps, unchecked nullable values, NonNullable casts without guards
Async errors: missing await, unhandled promise rejections, race conditions
Stale data: old values in caches, persisted state, or UI that survive restarts
Off-by-one: index boundaries, array length vs last index, loop bounds
Import errors: missing imports, circular dependencies, wrong paths
Security: unsanitized input, eval usage, SQL/command injection, hardcoded secrets
Error handling: swallowed exceptions, missing catch blocks, generic error responses

3. Build verification

Discover and run the project's build command:

Check package.json scripts, Makefile, Cargo.toml, pyproject.toml, or README
Run the build — confirm it succeeds
If multiple build targets exist (e.g., plugin + shared types), rebuild all

4. Type and test verification

Run the type checker (e.g., npx tsc --noEmit, pyright, cargo check)
Run the test suite (e.g., npm test, pytest, cargo test)
Fix all failures before proceeding
If no type checker or test suite exists: state this explicitly in the report

5. Architectural check

Find 2-3 examples of the same pattern used elsewhere in the codebase. Confirm the change is consistent with them. Flag if:

The change introduces a pattern not used anywhere else
A simpler approach already exists in the codebase
Unnecessary abstractions or over-engineering were added

6. Report

Present findings in this format:

## Review-Gate Report

| Check | Status | Notes |
|-------|--------|-------|
| Code path trace | PASS/FAIL | [what was traced] |
| Bug pattern scan | PASS/FAIL | [issues found or clean] |
| Build | PASS/FAIL | [command run + result] |
| Type check | PASS/FAIL/SKIP | [command run + result] |
| Tests | PASS/FAIL/SKIP | [command run + result] |
| Architecture | PASS/FAIL | [consistency notes] |

### Issues Found
- [list each issue and the fix applied, or "None"]

### Final Status: PASS / FAIL

Do not declare done until all checks PASS or are explicitly SKIP with reason.

What ships with it

Read from the repository

Just SKILL.md. No reference files, no scripts.