In the Skill's Own Terms
Reviews and improves a design plan before implementation. It rates design completeness from 0-10, explains the gaps, edits the plan and re-rates until the user accepts the result. The entry delegates seven detailed review passes to sections/review-sections.md. Mockups are the default for UI work when the designer is available, unless the user requests text-only review.
Includes an optional outside-voices step: Codex evaluates against OpenAI's design hard rules plus litmus checks; Claude subagent runs an independent completeness review. Operates in plan mode. For live site visual audits, use /design-review instead. The output is a better plan with a structured review report, not a document about the plan.
What it produces
- An improved plan file with design gaps closed, rated dimensions, and a structured GSTACK REVIEW REPORT section appended
- When mockups run: an HTML comparison board, feedback records and approved.json documenting the selected direction
- A litmus scorecard synthesizing Codex and Claude subagent findings if outside voices ran
How It Works
- 01Scope Gate Confirms Review Target
Before anything else, asks the user to confirm what to review (branch diff, plan doc, or specific page) unless plan mode auto-selects the active plan or the user explicitly named a target.
- 02Assess Scope, Then Preview
Audit project context and UI scope, rate the initial plan, and wait for the user’s focus-area response. Then generate mockups when the designer is available, unless the user declines. Complete visual selection and confirm the feedback before review passes.
- 03Optional Outside Design Voices Run in Parallel
If user opts in, launches Codex (design hard rules + litmus checks) and Claude subagent (independent completeness review) simultaneously.
- 04Seven Design Passes Rate 0-10 and Fix Gaps
Read the companion review-sections.md for the seven-pass workflow. Rate each dimension, explain the missing decisions, edit the plan, and re-rate until the target is met or the user chooses to move on. Scores describe the review rubric, not measured usability.
- 05Writes Structured Review Report to Plan File
Appends a GSTACK REVIEW REPORT section with runs/status/findings table, a VERDICT line, and an unresolved-decisions status as the file's terminal non-whitespace line.