clawpage-skill（抓爪技能）

Clawpage Skill (Router)

Purpose

This router only decides which sub-skill to invoke. All execution details (workflow, output, localization, checks, failure handling) are defined in sub-skills and shared contracts.

Sub-skills

1. init

• Path: ${CLAUDE_SKILL_DIR}/init/SKILL.md
• Purpose: initialize the skill, automatically register a new user, and save configuration to keys.local.json

1. create page

• Path: ${CLAUDE_SKILL_DIR}/create-page/SKILL.md
• Purpose: create a new page project and publish

1. update page

• Path: ${CLAUDE_SKILL_DIR}/update-page/SKILL.md
• Purpose: update an existing page project and republish

1. create management page

• Path: ${CLAUDE_SKILL_DIR}/create-management-page/SKILL.md
• Purpose: create or update the current read-only management page that lists user's pages

1. create template

• Path: ${CLAUDE_SKILL_DIR}/create-template/SKILL.md
• Purpose: create a reusable template folder

1. update template

• Path: ${CLAUDE_SKILL_DIR}/update-template/SKILL.md
• Purpose: update an existing template structure/style/interaction/docs

1. use-sdk

• Path: ${CLAUDE_SKILL_DIR}/use-sdk/SKILL.md
• Purpose: embed the Clawpage Browser SDK in a page (https://clawpage.ai/sdk.js) for live data, links, stats, blobs, /api/me

1. manage-data

• Path: ${CLAUDE_SKILL_DIR}/manage-data/SKILL.md
• Purpose: CRUD per-user KV data tables (comments / counters / configs / CMS) via the Clawpage data API

1. manage-links

• Path: ${CLAUDE_SKILL_DIR}/manage-links/SKILL.md
• Purpose: create / list / update / delete short links clawpage.ai/s/<slug> → *.clawpage.ai

1. manage-blobs

• Path: ${CLAUDE_SKILL_DIR}/manage-blobs/SKILL.md
• Purpose: upload images / files to Cloudflare R2 storage, get public blob.clawpage.ai/<id> URLs

1. view-stats

• Path: ${CLAUDE_SKILL_DIR}/view-stats/SKILL.md
• Purpose: page-view counts and per-day series for the user's pages / homepage / short links

Routing Priority (Conflict Resolution)

Apply this priority order when intent is mixed:

1. Initialization intent ("init", "setup", "初始化", "完成注册") -> init
2. Management-page intent ("管理页", "后台页", "列出我所有页面", "pages dashboard", "admin/read-only page list") -> create management page
3. Explicit page-id / pageId / "update existing page" signal -> update page
4. Existing local project intent (~/.clawpage/pages/<name>, ./.pages/<name>, "基于旧页面", "沿用现有页面") -> update page
5. Template-only intent (create/update template) -> create template or update template
6. Data / KV intent ("comments", "counters", "持久化", "data table") -> manage-data
7. Short-link intent ("short link", "缩短", "redirect", clawpage.ai/s/...) -> manage-links
8. Blob / image upload intent ("upload image", "上传图片", "PDF", "blob") -> manage-blobs
9. Stats intent ("view count", "traffic", "访问量", "stats") -> view-stats
10. SDK intent ("use SDK", "live data on page", "interactive page", "browser SDK") -> use-sdk
11. Otherwise default to creating a new page -> create page

Keyword Hints

• Init: "init", "setup", "初始化", "自动注册", "register"
• Create page: "new/create page", "from template", "发布新页面"
• Update page: "update/rework/revise", "existing page", "page-id"
• Create management page: "管理页", "页面管理", "列出所有页面", "dashboard of my pages", "read-only admin page"
• Create template: "new template", "模板搭建"
• Update template: "improve template", "模板改版"

Global Non-Negotiable Constraints

• Never remove required HTML placeholders: __CONTENT_HTML__, __DEFAULT_CSS__, __DEFAULT_JS__.
• Do not fabricate pageId for updates.
• Use API default https://api.clawpage.ai unless user overrides.
• For newly created pages, default publish policy is private + 3h TTL (pagecode required, ttlMs=10800000) unless user explicitly requests otherwise.
• Management page must be read-only (no destructive operations).
• Never write into the plugin install tree (${CLAUDE_SKILL_DIR}/...). Page projects go under ~/.clawpage/pages/<name>/ (the global default that @clawpage.ai/cli resolves bare names to). Treat plugin assets as read-only — they ship via npm and never need user-side edits.

References

• API semantics: ${CLAUDE_SKILL_DIR}/references/api-quickref.md
• Shared prompt contracts (output/localization/checks/errors): ${CLAUDE_SKILL_DIR}/references/prompt-contracts.md
• Publish entrypoint: npx -y @clawpage.ai/cli publish (npm package @clawpage.ai/cli)

Path Conventions

• Bash commands — invoke runtime via npx -y @clawpage.ai/cli <subcommand>. The CLI ships scripts + templates from npm; no plugin-relative paths needed, no permission prompts on ~/.claude/....
• Read-only references — ${CLAUDE_SKILL_DIR}/references/... for plugin-bundled docs (api / contracts / design guidelines). Read by Claude, not executed.
• Pages — [PAGE_DIR] resolved once per task. Default ~/.clawpage/pages/<name>/ (global workspace, since @clawpage.ai/cli 0.2.0). User may opt into project-scoped paths like ./.pages/<name>/ to check pages into a specific repo.
• Auth — ~/.clawpage/keys.local.json (auto-created by npx -y @clawpage.ai/cli init). Project-scoped ./keys.local.json in cwd takes precedence if present.