name: aiagent-lesson-runner description: "ai-agent-campのレッスンをCodexで開始・進行するスキル。 「レッスン開始」「次のレッスン」「start-0-1を始めたい」「Codexでレッスン」「スラッシュコマンドのレッスン」等のリクエストで発動。" version: 1.0.0 author: AI Brain Partners dependencies: [] triggers:

• aiagent-lesson-runner
• レッスン開始
• 次のレッスン
• レッスン実行
• Codexレッスン
• start-0-1
• lesson runner

トリガーワード

「レッスン開始」「次のレッスン」「レッスン実行」「スラッシュコマンド」「Codexレッスン」

AI Agent Lesson Runner

Use this skill to reproduce the start-* workflow in Codex.

Inputs

• A lesson id such as start-0-1

Workflow

1. Before running any lesson on a freshly cloned repo, run python3 tools/scripts/verify_integrity.py. If it reports the origin is not official (github.com/minicoohei/ai-agent-camp) or required files are missing, stop and ask the user to review the diff against upstream/main.
2. Validate that the lesson id matches ^start-\d+-\d+$. Reject any other value.
3. Check that .cursor/commands/lesson/<lesson-id>.md exists before reading it.
4. Open .cursor/commands/lesson/<lesson-id>.md.
5. Extract the lesson goal, prerequisites, checkpoints, command references, and referenced files.
6. Resolve the curriculum source in this order:

• frontmatter chapter when it points to a courses/aiagent/**/chapter*.yaml
• sibling practice/ and final/ docs next to that chapter
• if chapter is missing, infer the best source from courses/lessons.manifest.yaml, the lesson URL, and the start-X-Y module number

1. Treat embedded AskQuestion / AskUserQuestion JSON as a conversation blueprint. In Codex, convert it into concise numbered or bulleted options in normal chat instead of pretending the Cursor UI exists.
2. For /setup-start, /setup-github, /check-setup, and similar setup flows, do not tell the user to run the Cursor slash command literally. Execute or describe the underlying checks and split GUI-required steps from AI-executable steps.
3. If the lesson touches Git, secrets, MCP, or external APIs, tell the user which safety doc to read first and verify prerequisites before attempting the task.
4. Guide the user through:

• prerequisite check
• files to read
• actions to perform
• completion criteria
• next recommended lesson

Required References

• .cursor/commands/lesson/start-*.md
• matching courses/aiagent/**/chapter*.yaml when available
• sibling practice/ or final/ docs for the resolved chapter
• courses/lessons.manifest.yaml as the fallback lookup table when a lesson markdown is missing chapter

Safety

• If the lesson implies Git or environment changes, also consult docs/codex-safety.md.
• Never pretend the user can run the Cursor markdown command file directly in Codex.
• Never ask the user to paste secret values into chat. Reuse the repo credential workflow instead.

Expected Output

• Lesson summary
• Ordered next actions
• Relevant files
• Done criteria
• Suggested follow-up lesson
• When the lesson contains structured choices, present the Codex-friendly options inline in the response