メニュー
Run a full automated end-to-end test of the page-claw pipeline using a bundled fixture page-story. Simulates user responses autonomously. Use when you want to verify page-claw behavior after making changes to the skill.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
Batch generate static HTML page templates by combining a style matrix (design intents + layouts) with page-story markdown files. Skips interactive design questions — each style entry provides a design intent description that drives the full pipeline. Input: style-matrix.yaml + page-story files. Output: organized HTML + design docs.
Run an automated end-to-end test of the pageclaw-batch pipeline using bundled fixtures. Validates batch generation with a small style matrix. Use to verify pageclaw-batch behavior after changes.
Use when the user wants to turn a page-story markdown file (page-story-*.md) into a polished static HTML page. Trigger for: personal pages, academic homepages, portfolio pages, profile pages, any request to build a page from a structured markdown brief. Runs the full pipeline automatically: design context gathering, design system, implementation plan, build, and quality pass.
One-time setup that gathers design context for your project and saves it to your AI config file. Run once to establish persistent design guidelines.
| name | pageclaw-test |
| description | Run a full automated end-to-end test of the page-claw pipeline using a bundled fixture page-story. Simulates user responses autonomously. Use when you want to verify page-claw behavior after making changes to the skill. |
| license | MIT |
| metadata | {"author":"pageclaw","version":"1.0.0","homepage":"https://github.com/XY-Showing/pageclaw"} |
Run the full page-claw pipeline on the bundled fixture page-story, simulating user responses autonomously, and produce a test report.
Page-story location: skills/pageclaw-test/fixtures/page-story-test.md (relative to the pageclaw repo root, or the equivalent installed path under ~/.claude/skills/pageclaw-test/fixtures/).
Read this file at the start. This is the page-story for the entire test run.
When page-claw asks questions, respond as a simulated user using these rules — do not break character or explain what you're doing until the pipeline is complete:
| Question type | Response strategy |
|---|---|
| Multiple choice (options provided) | Pick one option at random |
| Open-ended (no options given) | Ask: "What are my options?" — wait for options, then pick one at random |
| Skip/optional (e.g. reference design) | Always respond: "Skip" |
| Any question beyond the expected flow (see Behavior Checks) | Record it as a violation and respond: "Skip" |
Save all pipeline artifacts to test-output/ in the current working directory:
test-output/design.md — design context + design system (Steps 1–2)test-output/impl.md — implementation plan (Step 3)test-output/index.html — final page (Step 4)Do not save to docs/plans/ — this is a test run, not a real project.
After the pipeline completes, print a test report in this format:
## Page Claw Test Report
### Pipeline
- [ ] Step 1 — Design Context: completed / FAILED
- [ ] Step 2 — Design System: completed / FAILED
- [ ] Step 3 — Implementation Plan: completed / FAILED
- [ ] Step 4 — Build: completed / FAILED
- [ ] Step 5 — Quality pass: completed / skipped / FAILED
### Behavior Checks
- [ ] Q1 was skip-optional (reference design)
- [ ] Q2 was multiple choice with ≥3 options (visual direction) — only present if Q1 was skipped
- [ ] Q3 was multiple choice with 4 options (aesthetic style) — always present
- [ ] No additional questions asked beyond Q1 + Q2 (if applicable) + Q3
- [ ] No references to existing project files in design decisions
### Violations
(list any violations found, or "None")
### Output
- design.md: saved / missing
- impl.md: saved / missing
- index.html: saved / missing
### Notes
(any observations about output quality, faithfulness to page-story, etc.)
Mark each item as passed (check) or failed (with a brief note). After the report, open test-output/index.html in the browser if possible.