The planner work node workflow (w11z1assa) completed and produced a BRD for a bug fix: the `#/skills/:skill_id/audit` detail page shows audit/iteration history but never displays the skill's actual `SKILL.md` content.

Key points:
- Root cause: `src/skill_query.py` never had a function to read `skills/<skill_id>/SKILL.md` from disk — only DB-backed `list_skills`/`get_skill_audit_log`. This gap existed since U-029 but only became user-visible once real skill files were migrated into `skills/`.
- Scope in: new `skill_query` function to read skill content + list supporting files (references/templates/scripts), a new read-only `GET /api/skills/:skill_id/content` route, a new "Skill 内容" panel in `frontend/app.js` alongside the existing iteration-history panel, plus backend unit tests (normal read, missing skill, missing file, path-traversal rejection).
- Scope out: no changes to `skill_update.py` migration logic, no per-file browsing of support files, no markdown rendering/highlighting, no changes to the `#/skills` list page or existing audit/diff/rollback behavior.
- Notable risk: first case in the repo of a URL path segment feeding into filesystem path construction — explicit path-traversal protection called out as new work.

Full BRD text (truncated in the notification) is saved at /private/tmp/claude-501/-Users-yibo-Documents-lemonAI-graphlemon--graphlemon-worktrees-33e0266c-d2ec-4982-9128-afd2d36aea4d/21d170c6-531d-4f7e-8af1-b88e06014cf3/tasks/w11z1assa.output. Awaiting user direction on whether to proceed to PRD/SDD or hand off to the Coder.