菜单
Verify a CodeXR change before declaring it done — route to the narrowest test script for the touched area, run the npm test gate, and drive the F5/browser flow when a runtime surface changed. Use whenever a code change is about to be reported as working.
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/aMonteSl/CodeXR --skill verify命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
Add or adjust supported-language coverage in CodeXR — the Python metric contract, TS language metadata, dependency-graph support, fixtures, and the analysis test matrix. Use when adding a language, changing per-language metric capabilities, or investigating language-count discrepancies.
Capture the launch screenshots for the README and the website (LivePanel pages, XR scenes, guide.html) by driving a real running analysis with Playwright, reviewing every shot before keeping it. Use when asked to produce, retake or optimize any asset listed in media/SHOTLIST.md.
Add or modify a browser-side runtime under templates/components/ (XR scene runtimes, LivePanel shared components) — injection path, load order, bundling, and the required tests. Use when touching any JS/CSS under templates/components/ or wiring a new component asset.
基于 SOC 职业分类
正在显示 SKILL.md
| name | verify |
| description | Verify a CodeXR change before declaring it done — route to the narrowest test script for the touched area, run the npm test gate, and drive the F5/browser flow when a runtime surface changed. Use whenever a code change is about to be reported as working. |
Never claim a change works without running tests or the F5 flow. Route by what was touched, run the narrow script first (faster failure), then finish with the default gate.
| Touched | Run first |
|---|---|
src/code_analysis/python/** | npm run test:python, then npm run test:analysis if output schema/metrics changed |
python/html/html_dom_parser.py or DOM mode | npm run test:htmlanalysis |
templates/components/** runtimes | the matching test/analysis/*.test.cjs via node --test test/analysis/<suite>.test.cjs; table/containment → npm run test:xr-harness; project evolution → npm run test:project-evolution-harness |
templates/analysis_livePanel/** | node --test test/analysis/directoryLivePanelUi.test.cjs test/analysis/livePanelDomIntegration.test.cjs |
src/servers/** | suites under test/servers/ |
src/commands/** or command registration | node --test test/commands/commandRegistry.test.cjs (unique IDs enforced) |
| Language support / metrics output | npm run test:analysis (XR and LivePanel must produce identical data.json) |
Single suite: node --test --test-isolation=none --test-concurrency=1 test/<area>/<file>.test.cjs. Unit tests are Node native test runner .test.cjs files — not Mocha. Some suites assert against compiled output: run npm run compile-tests (tsc → out/) first if a suite imports from out/.
npm test # typecheck + lint + compile-tests + test:unit
npm run compile # webpack build, catches template/asset copy issues
Both must be green. Report the actual counts (e.g. "339/339").
If the change affects generated scenes, LivePanel pages, or servers, automated tests are not sufficient evidence on their own:
.vscode-test/dev-extension) → open a workspace → right-click file/folder → "Code-XR: Analysis" → run the affected mode (LivePanel / XR / DOM).docs/xr-testing/XR_DEBUG_COMMANDS.md (CodeXR.help()).confirmed (you ran it) / self-assessed (a doc claims it) / inferred / open. Never upgrade a label to make the report look better.