SKILL.md

AgentClick Skill Router

Use this file only to route to the right sub-skill.

Base flow:

1. Detect AgentClick URL: try http://localhost:38173/api/health first; if unreachable, use http://host.docker.internal:38173.
2. If the server is not reachable at either address, start AgentClick locally with npm start, then re-check.
3. Create a session with POST /api/review.
4. Monitor the session — polling strategy depends on environment (see sub-skill for details):
   • Non-Docker (Claude Code, Codex, local): use GET /api/sessions/:id/wait (blocks until state changes).
   • Docker (OpenClaw): use short-poll GET /api/sessions/:id one call per exec; you are the loop controller.
5. If status is rewriting, update with PUT /api/sessions/:id/payload and continue monitoring the same session until status=completed or pageStatus.stopMonitoring=true.

Sub-skills:

• action_approval -> skills/clickui-approve/SKILL.md
• code_review -> skills/clickui-code/SKILL.md
• email_review -> skills/clickui-email/SKILL.md
• plan_review -> skills/clickui-plan/SKILL.md
• trajectory_review -> skills/clickui-trajectory/SKILL.md
• memory_review and memory_management -> skills/clickui-memory/SKILL.md

Keyword routing for UI review:

• action, approval -> action_approval
• command, shell, script, diff, code -> code_review
• email, draft, reply -> email_review
• plan, steps, strategy -> plan_review
• trajectory, run log -> trajectory_review
• memory, memory files, browse memory -> skills/clickui-memory/SKILL.md