fix skill dispatch ref: pr-review is a command not a skill

scripts/generate-skill-index.py

@@ -430,6 +430,8 @@ def _process_skill_dir(skill_dir: Path, skill_file_override: Path | None = None)
             old_ok = (repo_root / existing["file"]).is_file() if repo_root else False
             if old_ok and not new_ok:
                 return
+            if old_ok and new_ok:
+                warnings.append(f"  - {name}: duplicate skill name — {existing['file']} overwritten by {entry['file']}")
         index[collection_key][name] = entry
 
     for child in sorted(source_dir.iterdir()):
@@ -443,6 +445,16 @@ def _process_skill_dir(skill_dir: Path, skill_file_override: Path | None = None)
         # Check if this directory directly contains a SKILL.md (flat layout)
         if (child / "SKILL.md").exists():
             _process_skill_dir(child)
+            # Hybrid dirs (e.g. skills/process/, skills/research/) have both a top-level SKILL.md and nested skill subdirs.
+            for nested in sorted(child.iterdir()):
+                if not nested.is_dir():
+                    continue
+                if nested.is_symlink() and not include_private:
+                    continue
+                if (nested / "SKILL.md").exists():
+                    _process_skill_dir(nested)
+                elif (nested / "skill" / "SKILL.md").exists():
+                    _process_skill_dir(nested, skill_file_override=nested / "skill" / "SKILL.md")
         # Check for nested skill/SKILL.md layout (e.g., voice-example/skill/SKILL.md)
         elif (child / "skill" / "SKILL.md").exists():
             _process_skill_dir(child, skill_file_override=child / "skill" / "SKILL.md")

scripts/tests/test_generate_skill_index.py

@@ -376,3 +376,56 @@ def test_include_private_run_writes_only_the_requested_output(self, tmp_path: Pa
         data = json.loads(output.read_text(encoding="utf-8"))
         # Floor tracks the post-consolidation catalog; raise it with new skills.
         assert 35 <= len(data["skills"]) <= 80
+
+
+class TestHybridDirIndexing:
+    """Hybrid dirs (top-level SKILL.md + nested skill subdirs) appear fully in index."""
+
+    def test_hybrid_dir_top_and_nested_both_indexed(self, tmp_path: Path) -> None:
+        """A hybrid dir's top-level skill and its nested skill both appear in the index."""
+        skills_dir = tmp_path / "skills"
+        # Top-level hybrid skill (e.g. skills/process/SKILL.md)
+        process_dir = skills_dir / "process"
+        process_dir.mkdir(parents=True)
+        (process_dir / "SKILL.md").write_text(_SKILL_FRONTMATTER.format(name="process"))
+        # Nested skill under the hybrid dir (e.g. skills/process/pr-workflow/SKILL.md)
+        nested_dir = process_dir / "pr-workflow"
+        nested_dir.mkdir()
+        (nested_dir / "SKILL.md").write_text(_SKILL_FRONTMATTER.format(name="pr-workflow"))
+
+        index, _warnings = gsi.generate_index(
+            source_dir=skills_dir,
+            dir_prefix="skills",
+            collection_key="skills",
+        )
+
+        assert "process" in index["skills"], "Top-level hybrid skill must be indexed"
+        assert "pr-workflow" in index["skills"], "Nested skill in hybrid dir must be indexed"
+
+
+class TestOverwriteWarning:
+    """A duplicate skill name emits an overwrite warning."""
+
+    def test_duplicate_name_emits_warning(self, tmp_path: Path) -> None:
+        """Two skill dirs with the same name: in frontmatter triggers the overwrite-warning branch."""
+        skills_dir = tmp_path / "skills"
+        # Two different dir names, same name: in frontmatter so both map to "shared-name".
+        dir_a = skills_dir / "cat-a" / "shared-skill"
+        dir_a.mkdir(parents=True)
+        (dir_a / "SKILL.md").write_text(_SKILL_FRONTMATTER.format(name="shared-name"))
+
+        dir_b = skills_dir / "cat-b" / "shared-skill"
+        dir_b.mkdir(parents=True)
+        (dir_b / "SKILL.md").write_text(_SKILL_FRONTMATTER.format(name="shared-name"))
+
+        index, warnings = gsi.generate_index(
+            source_dir=skills_dir,
+            dir_prefix="skills",
+            collection_key="skills",
+            repo_root=tmp_path,
+        )
+
+        assert "shared-name" in index["skills"], "The skill must still be indexed"
+        assert any("shared-name" in w for w in warnings), (
+            f"Expected an overwrite warning for 'shared-name'; got warnings: {warnings}"
+        )

scripts/validate-skill-names.py

@@ -38,6 +38,12 @@ def find_duplicates(repo: Path) -> dict[str, list[str]]:
             continue
         if (top / "SKILL.md").is_file():
             seen.setdefault(top.name, []).append(top.relative_to(repo).as_posix())
+            # Hybrid dir: also scan immediate subdirs for nested skills.
+            # Intentionally scans only 1 level deep (vs. generate-skill-index.py's 2 levels):
+            # this is a fast-fail gate, not an exhaustive mirror of the indexer.
+            for child in sorted(top.iterdir()):
+                if child.is_dir() and (child / "SKILL.md").is_file():
+                    seen.setdefault(child.name, []).append(child.relative_to(repo).as_posix())
             continue
         for child in sorted(top.iterdir()):
             if child.is_dir() and (child / "SKILL.md").is_file():

skills/meta/toolkit/references/toolkit-evolution.md

@@ -238,7 +238,7 @@ Win condition for each implementation:
 
 **Step 1: Handle winners (WIN status)**
 
-For each winning implementation, create a PR using the template from `references/evolve-scripts.md` § Step 1, then merge. After creating the PR, run pr-review to validate, then merge.
+For each winning implementation, create a PR using the template from `references/evolve-scripts.md` § Step 1, then merge. After creating the PR, run /pr-review to validate, then merge.
 
 The multi-persona critique + A/B testing gate is the review. Auto-merge is safe because the validation happened before this step.
 

skills/process/pr-workflow/references/fix.md

@@ -206,8 +206,8 @@ Solution:
 
 ### Related Skills
 - `pr-pipeline` — For creating PRs from scratch
-- `pr-review` — For reviewing code without fixing
-- `systematic-debugging` — For general debugging unrelated to PR comments
+- `review` — For reviewing code without fixing (skill); use `/pr-review` command for full multi-agent PR review
+- `debugging` — For general debugging unrelated to PR comments
 
 ### PR Comment Best Practices
 - Always validate before fixing (this prevents introducing bugs that reviewers caught)

skills/process/pr-workflow/references/sync.md

@@ -221,7 +221,7 @@ Solution:
 
 ## References
 
-- `/pr-review` -- Comprehensive PR review (used in the review-fix loop)
+- `/pr-review` -- Comprehensive PR review slash command (not the `review` skill; used in the review-fix loop)
 - `/pr-cleanup` -- Post-merge branch cleanup
 - `scripts/classify-repo.py` -- Repo classification for workflow gating
 - `scripts/adr-decision-coverage.py` -- ADR decision coverage checker