Claude-skill-registry implementation-review
Use after implementing a plan to verify completeness, correctness, and merge safety - the post-implementation gate
install
source · Clone the upstream repo
git clone https://github.com/majiayu000/claude-skill-registry
Claude Code · Install into ~/.claude/skills/
T=$(mktemp -d) && git clone --depth=1 https://github.com/majiayu000/claude-skill-registry "$T" && mkdir -p ~/.claude/skills && cp -r "$T/skills/data/implementation-review" ~/.claude/skills/majiayu000-claude-skill-registry-implementation-review && rm -rf "$T"
manifest:
skills/data/implementation-review/SKILL.mdsource content
Implementation Review
Overview
Verify implementation is complete, correct, and safe to merge. This is the merge readiness gate.
Core principle: Catch issues before merge, not after.
Announce at start: "I'm using the implementation-review skill to audit this implementation."
Plan location:
plans/active/{plan-name}/
The Process
Step 1: Load Context
- Read original plan from
plans/active/{plan-name}/ - Get current branch name
- Find base branch (main/master)
git branch --show-current git merge-base HEAD main 2>/dev/null || git merge-base HEAD master
Step 2: Gather Implementation Data
Get all commits since diverging from base:
git log --oneline $(git merge-base HEAD main)..HEAD
Get all changed files:
git diff --name-only $(git merge-base HEAD main)..HEAD
Step 3: Evaluate Against Criteria
Correctness:
- Implementation matches plan intent?
- Any drift from plan (features added/removed)?
- All tasks from plan completed?
Completeness:
- Success criteria from plan met?
- All expected files exist?
- Nothing half-done or TODO'd?
Quality:
- Code follows project patterns?
- No obvious bugs?
- Error handling present where needed?
Safety:
- Tests passing?
- No regressions in existing tests?
- No security issues introduced?
- No debug code left (console.log, print, etc.)?
Step 4: Run Checks
Run test suite:
# Detect and run appropriate test command npm test || cargo test || pytest || go test ./...
Check for debug code:
git diff $(git merge-base HEAD main)..HEAD | grep -E "(console\.log|print\(|debugger|TODO|FIXME)" || echo "None found"
Step 5: Produce Verdict
Output this format:
## Implementation Review: {plan-name} ### Verdict: MERGE READY | NEEDS FIXES | MAJOR ISSUES ### Plan Compliance - [X] Task 1: [Description] - Implemented correctly - [X] Task 2: [Description] - Implemented correctly - [ ] Task 3: [Description] - Partial/Missing [reason] ### Test Results - Suite: {X}/{Y} passing - Coverage: {Z}% (if available) ### Quality Checks - [X] No debug code (console.log, print, etc.) - [X] Follows project patterns - [X] Error handling present ### Issues Found - **[Severity: blocker|major|minor]**: [Description] - Location: [file:line] - Suggested fix: [How to address] ### Recommendation [Ready to merge / Fix these N issues first]
Verdict Meanings
| Verdict | Meaning | Next Step |
|---|---|---|
| MERGE READY | Safe to merge | Finish branch (merge/PR) |
| NEEDS FIXES | Minor issues to address | Fix, re-review |
| MAJOR ISSUES | Significant problems | May need plan revision |
Issue Severity
| Severity | Meaning |
|---|---|
| blocker | Cannot merge until fixed (test failure, security issue) |
| major | Should fix before merge (missing functionality, bad pattern) |
| minor | Nice to fix but not blocking (style, minor improvements) |
Red Flags
Never:
- Approve with failing tests
- Skip debug code check
- Ignore plan drift without noting it
Always:
- Run actual test suite
- Compare against original plan
- Check for leftover debug artifacts
Integration
Invoked by:
- gremlins:worktree-workflow (audit phase) - Post-implementation gate
- Standalone for any implementation review
Follows:
- gremlins:executing-plans - Reviews output of implementation
Leads to:
- gremlins:finishing-a-development-branch - When MERGE READY
- gremlins:reach-opportunities - When user opts in after passing
vs verification-before-completion:
- Use implementation-review as the workflow gate (comprehensive, plan-aware)
- Use verification-before-completion for quick ad-hoc checks outside workflow