菜单
Drive the running ClawStash app in a real browser to verify UI/UX changes end-to-end. Use after nontrivial frontend changes, before committing — tests and tsc alone have missed lifecycle/timing bugs in this repo before (see MEMORY.md
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/fo0/clawstash --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。
Orchestrator mode — the default working mode of this project, not a toggle to find. The main agent does no task work itself: every unit goes to a role-framed subagent that inherits its model and effort, at most 5 in parallel unless overridden. Load it for the contract (roles, width, write scopes, verification), when the user says '/orca', 'orca mode', 'orchestrator mode', 'orca an/aus', 'delegate everything', or asks what the width is set to. '/orca off' drops to plain behavior for this session only.
Use when the user wants CI status, failed-job logs, or help fixing a red build. Triggered by /ci, 'CI status', 'check the build', 'fix CI', 'why is CI failing', 'look at the build'. Auto-routes by state: status / logs / fix-proposal. Reads logs locally — never re-triggers builds without explicit user command.
Use for any GitHub Pull Request work. Auto-detects lifecycle phase (create / update / report) from current state — only requires explicit command for status, comments, or merge. Triggered by /pr, 'PR', 'create PR', 'open PR', 'update PR', 'PR status', 'merge PR'. Suggests, never auto-creates without user invocation.
基于 SOC 职业分类
正在显示 SKILL.md
| name | verify |
| description | Drive the running ClawStash app in a real browser to verify UI/UX changes end-to-end. Use after nontrivial frontend changes, before committing — tests and tsc alone have missed lifecycle/timing bugs in this repo before (see MEMORY.md |
npm install # fresh containers start without node_modules
DATABASE_PATH=/tmp/verify.db PORT=3100 npm run dev # isolated scratch DB, non-default port
# server is up when: curl -s -o /dev/null -w "%{http_code}" http://localhost:3100/ -> 200 (~1-5 s)
Use a scratch DATABASE_PATH outside the repo — never let a verification run
create or touch ./data/.
Chromium is pre-installed for Playwright in remote sessions
(/opt/pw-browsers/chromium-*/chrome-linux/chrome; ls /opt/pw-browsers/).
The repo does not depend on Playwright — install playwright-core into a
scratch dir (NOT the repo) and point chromium.launch({ executablePath }) at
the pre-installed binary. Do not run playwright install.
Useful selectors/flows that exist today:
.btn-new-stash → #stash-name, .file-name-input,
.code-editor-wrapper textarea, button:has-text("Save Stash")./stash/<uuid>; editor: /stash/<uuid>/edit.? → .shortcuts-help-dialog, Alt+K → .search-overlay.n, e, a, Escape) are inert while typing in inputs —
to exercise them, focus a button/heading first.window.confirm — handle Playwright dialog events./graph, canvas .graph-canvas; hit-test nodes by probing a small
click grid around the canvas center.npm run format (pinned Prettier from node_modules) before judging
git status — hooks/other Prettier versions can reformat files you never
touched; the pinned run restores them.npm install with a different npm major may churn package-lock.json
(e.g. libc fields). If you added no dependency, git checkout -- package-lock.json.pkill -f "next dev").