From tac
Designs spec-based review workflows using screenshots for visual proof, issue severity classification, and auto-resolution loops to validate implementations.
npx claudepluginhub melodic-software/claude-code-plugins --plugin tacThis skill is limited to using the following tools:
Design review workflows that validate implementation against specifications with visual proof.
Reviews code changes against spec via git diffs and screenshots. Classifies deviations by severity and outputs JSON summary to verify post-implementation alignment.
Verifies implementation matches design specification for functional completeness, TDD compliance, and test coverage. Stage 1 of two-stage review.
Verifies code implementations match specs, PRDs, epics, or tasks by checking completeness, acceptance criteria, edge cases, and scope creep. Use post- or during-implementation.
Share bugs, ideas, or general feedback.
Design review workflows that validate implementation against specifications with visual proof.
Review answers: "Is what we built what we asked for?"
This is different from testing which answers: "Does it work?"
Establish where specifications live:
specs/
├── feature-{name}.md
├── bug-{name}.md
└── chore-{name}.md
Or:
specs/issue-{number}-{type}.md
Identify critical paths that need visual proof:
Example:
1. Take screenshot of dashboard (initial state)
2. Click "Export" button
3. Take screenshot of export modal
4. Complete export
5. Take screenshot of success message
Configure issue severity levels:
| Severity | Criteria | Action |
|---|---|---|
| blocker | Prevents release, harms UX | Auto-resolve |
| tech_debt | Quality issue, works | Document |
| skippable | Polish, preference | Ignore |
Design the auto-resolution loop:
Review
├── Issues found?
│ ├── Blockers? → /patch → /implement → Re-review
│ └── No blockers → Success
└── No issues → Success
Maximum 3 retry attempts
Options for screenshot storage:
agents/{adw_id}/review_img/# Review Implementation Against Spec
## Variables
spec_file: $1
## Instructions
1. **Read Specification**
- Understand requirements
- Note success criteria
2. **Analyze Changes**
- git diff origin/main
- Compare against spec
3. **Capture Screenshots**
- 1-5 screenshots of critical functionality
- Number: 01_name.png, 02_name.png
4. **Classify Issues**
- blocker: Must fix before release
- tech_debt: Document for later
- skippable: Can ignore
## Output Format
{
"success": boolean,
"review_summary": "string",
"review_issues": [...]
}
{
"issue_description": "What's wrong",
"issue_resolution": "How to fix it",
"issue_severity": "blocker| tech_debt |skippable"
}
for attempt in range(1, MAX_ATTEMPTS + 1):
review_result = run_review()
if review_result.success:
break # No blockers
blockers = filter(issues, severity="blocker")
for blocker in blockers:
create_patch(blocker)
implement_patch()
# Loop continues with re-review
/plan → What are we building?
/build → Make it real
/test → Does it work?
/review → Is it what we asked for? ← THIS SKILL
/patch → Fix blockers
/document → How does it work?
Date: 2025-12-26 Model: claude-opus-4-5-20251101