diff --git a/.graphlemon/runs/b243dfe8-4da8-49bd-895b-8b9e9ed17280/brd.md b/.graphlemon/runs/b243dfe8-4da8-49bd-895b-8b9e9ed17280/brd.md
new file mode 100644
index 0000000..37daf63
--- /dev/null
+++ b/.graphlemon/runs/b243dfe8-4da8-49bd-895b-8b9e9ed17280/brd.md
@@ -0,0 +1,9 @@
+The planner Work Node finished. It produced a BRD for fixing a bug where the Skill detail page (`#/skills/:skill_id/audit`) doesn't show a skill's actual content — only the iteration/audit log, which is often empty since `src/skill_query.py` never reads `SKILL.md` off disk.
+
+Key points from the BRD:
+- **Root cause**: `skill_query.py` has no function reading `skills/<skill_id>/SKILL.md`; only iteration-log and diff queries exist.
+- **Scope**: add a read API for skill content (with path-traversal protection on `skill_id`), show it alongside (not replacing) the existing iteration panel, plain `<pre>` text rendering (no markdown lib, matching `v_artifact_browser` convention), and list any supporting files (`references/`, `templates/`, `scripts/`).
+- **Non-goals**: no changes to migration logic, no edits to iteration-log display, no write/edit capability, no generic binary-file preview.
+- **Acceptance criteria**: 7 concrete checks including path-traversal rejection and zero regression in existing skill tests.
+
+This is BRD-stage output only — no code was changed. Next step per the doc is PRD/SDD refinement of two open design points (how to surface supporting files, and the new endpoint's shape).
\ No newline at end of file