Files

Ken Yasue a579d90113 Add design-first workflow aligned with opencode boilerplate

- New commands: /define-design, /update-design, /generate-app (web | flutter | winui3)
- New skills: ui-design, design-tokens, platform-ui-generation
- /add-feature: check design spec before implementation (Step 3.5) and reconcile all six persistent docs after implementation (Step 8)
- /setup-project: automatic document creation with self-check checkpoints
- CLAUDE.md: document docs/design/, .claude/ configuration, and the design phase in the development process
- Enrich subagent descriptions; allowlist new skills in settings.json
- Add docs/design/README.md

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

2026-07-02 09:00:03 +02:00

1.6 KiB

Raw Blame History

description
Run a detailed document review using a subagent

Document Review

Document path: $ARGUMENTS (e.g. /review-docs docs/product-requirements.md)

How to Run

claude
> /review-docs docs/product-requirements.md

Procedure

Step 1: Verify the Document Exists

Check whether the specified document exists.

Step 2: Launch the doc-reviewer Subagent

Launch the doc-reviewer subagent to run the review:

Use the Task tool to launch the doc-reviewer subagent:

subagent_type: "doc-reviewer"
description: "Document detailed review"
prompt: "Please review $ARGUMENTS in detail.\n\nEvaluate it from the following perspectives:\n1. Completeness: Are all required items included?\n2. Specificity: Are there any ambiguous expressions?\n3. Consistency: Is it consistent with other documents?\n4. Measurability: Are success metrics measurable? (for a PRD)\n\nPlease produce a review report."

Step 3: Summarize the Review Results

Extract the key points from the review report produced by the subagent and report them to the user.

Output Format

# Document Review Results

## Document: [file name]

### Key Improvements

1. [Improvement 1] (Priority: High/Medium/Low)
2. [Improvement 2] (Priority: High/Medium/Low)
3. [Improvement 3] (Priority: High/Medium/Low)

### Overall Rating

[1-5]/5

### Next Actions

- [Recommended action 1]
- [Recommended action 2]

For the full report, refer to the subagent's output.

Notes

The review involves detailed analysis and may take a few minutes.
The subagent runs in an independent context, so it does not consume the main agent's context.

1.6 KiB Raw Blame History