diff --git a/.claude/skills/bmad-os-audit-file-refs/SKILL.md b/.claude/skills/bmad-os-audit-file-refs/SKILL.md
new file mode 100644
index 000000000..484d5086c
--- /dev/null
+++ b/.claude/skills/bmad-os-audit-file-refs/SKILL.md
@@ -0,0 +1,7 @@
+---
+name: bmad-os-audit-file-refs
+description: Audit BMAD source files for file-reference convention violations using parallel Haiku subagents. Use when checking path references in workflow and task files.
+disable-model-invocation: true
+---
+
+Read `prompts/instructions.md` and execute.
diff --git a/tools/audit-file-refs.md b/.claude/skills/bmad-os-audit-file-refs/prompts/instructions.md
similarity index 100%
rename from tools/audit-file-refs.md
rename to .claude/skills/bmad-os-audit-file-refs/prompts/instructions.md
diff --git a/tools/maintainer/review-pr-README.md b/.claude/skills/bmad-os-review-pr/README.md
similarity index 87%
rename from tools/maintainer/review-pr-README.md
rename to .claude/skills/bmad-os-review-pr/README.md
index d097ce948..b5cc612d1 100644
--- a/tools/maintainer/review-pr-README.md
+++ b/.claude/skills/bmad-os-review-pr/README.md
@@ -6,9 +6,9 @@ Adversarial code review for GitHub PRs. Works with any LLM agent.
## How It Works
-Point your agent at `review-pr.md` and ask it to review a specific PR:
+Use `/bmad-os-review-pr` to review a specific PR:
-> "Read tools/maintainer/review-pr.md and apply it to PR #123"
+> "Use /bmad-os-review-pr to review PR #123"
The tool will:
@@ -17,7 +17,7 @@ The tool will:
3. Transform findings into professional tone
4. Preview the review and ask before posting
-See `review-pr.md` for full prompt structure, severity ratings, and sandboxing rules.
+See `prompts/instructions.md` for full prompt structure, severity ratings, and sandboxing rules.
## When to Use
diff --git a/.claude/skills/bmad-os-review-pr/SKILL.md b/.claude/skills/bmad-os-review-pr/SKILL.md
new file mode 100644
index 000000000..bb8925674
--- /dev/null
+++ b/.claude/skills/bmad-os-review-pr/SKILL.md
@@ -0,0 +1,7 @@
+---
+name: bmad-os-review-pr
+description: Adversarial PR review tool (Raven's Verdict). Cynical deep review transformed into professional engineering findings. Use when asked to review a PR.
+disable-model-invocation: true
+---
+
+Read `prompts/instructions.md` and execute.
diff --git a/tools/maintainer/review-pr.md b/.claude/skills/bmad-os-review-pr/prompts/instructions.md
similarity index 96%
rename from tools/maintainer/review-pr.md
rename to .claude/skills/bmad-os-review-pr/prompts/instructions.md
index 24dbb7069..12d150049 100644
--- a/tools/maintainer/review-pr.md
+++ b/.claude/skills/bmad-os-review-pr/prompts/instructions.md
@@ -2,8 +2,7 @@
A cynical adversarial review, transformed into cold engineering professionalism.
-
-CRITICAL: Sandboxed Execution Rules
+## CRITICAL: Sandboxed Execution Rules
Before proceeding, you MUST verify:
@@ -14,9 +13,6 @@ Before proceeding, you MUST verify:
**If no explicit PR number/URL was provided, STOP immediately and ask:**
"What PR number or URL should I review?"
-
-
-
## Preflight Checks
@@ -97,9 +93,7 @@ gh pr diff {PR_NUMBER} [--repo {REPO}] --name-only | grep -E '\.(png|jpg|jpeg|gi
Store list of binary files to skip. Note them in final output.
-
-
-
+## Adversarial Review
### 1.1 Run Cynical Review
@@ -130,9 +124,7 @@ Likely tag:
- Add `[likely]` to findings with high confidence, e.g. with direct evidence
- Sort findings by severity (Critical → Moderate → Minor), not by confidence
-
-
-
+## Tone Transformation
**Transform the cynical output into cold engineering professionalism.**
@@ -177,9 +169,8 @@ Output format after transformation:
_Review generated by Raven's Verdict. LLM-produced analysis - findings may be incorrect or lack context. Verify before acting._
```
-
+## Post Review
-
### 3.1 Preview
Display the complete transformed review to the user.
@@ -231,12 +222,10 @@ Do NOT use heredocs or `echo` - Markdown code blocks will break shell parsing. U
Keep the temp file and inform user of location.
-
+## Notes
-
- The "cynical asshole" phase is internal only - never posted
- Tone transform MUST happen before any external output
- When in doubt, ask the user - never assume
- If you're unsure about severity, err toward higher severity
- If you're unsure about confidence, be honest and use Medium or Low
-