course-recommendation-agent

name	course-recommendation-agent
description	Use this skill when the user wants coding course recommendations powered by the local course-agent service, including online search, preference-aware ranking, and a confirmation step before the final recommendation is generated.

Course Recommendation Agent

Use this skill when the user asks for code-learning recommendations such as Python 入门, JavaScript 前端课程, LangChain 实战, 算法课推荐, or 适合转行的后端课程.

This skill operates the local FastAPI service in this repository instead of recreating the recommendation logic in-place.

Workflow

1. Ensure the local service is available.
2. If needed, start the service with scripts/start_server.ps1.
3. Check readiness with scripts/health_check.ps1.
4. Send the initial request with scripts/chat.ps1.
5. If the response status is needs_confirmation, show the candidate courses to the user and ask whether to continue.
6. If the user confirms, resume with scripts/confirm.ps1 -Action confirm.
7. If the user cancels, resume with scripts/confirm.ps1 -Action cancel.

Rules

• Do not treat needs_confirmation as the final answer.
• Preserve course source links in any summary you present.
• Pass through the user's stated preferences for language, platform, style, and goals.
• Prefer reusing the existing session_id for follow-up requests in the same conversation.
• If the local service cannot start or health checks fail, report that clearly instead of fabricating results.

Scripts

• scripts/start_server.ps1: starts the FastAPI app with the repo virtual environment.
• scripts/health_check.ps1: checks /health.
• scripts/chat.ps1: sends an initial recommendation request to /api/chat.
• scripts/confirm.ps1: resumes a paused workflow through /api/chat/confirmation.

References

• Read references/api.md for request and response shapes.
• Read references/workflow.md for the confirmation lifecycle and recommended operator behavior.