diff --git a/src/api/schemas.py b/src/api/schemas.py
index e2e5fba..df5211e 100644
--- a/src/api/schemas.py
+++ b/src/api/schemas.py
@@ -1597,9 +1597,20 @@ class SkillCatalogItem(BaseModel):
     )
 
 
-class SkillSuggestionRequest(ProjectScopedRequest):
+class SkillSuggestionRequest(BaseModel):
     """Request to suggest skills for a role in a project."""
 
+    model_config = ConfigDict(populate_by_name=True)
+
+    project_id: ProjectId | None = Field(
+        alias="projectId",
+        default=None,
+        description=(
+            "Optional project this request is scoped to. When absent, retrieval "
+            "is skipped and only universal/role-typical skills are suggested."
+        ),
+        examples=["3f1c0b1e-1f4d-4a5e-9b6a-0d2c8f7e5a11"],
+    )
     role_name: str = Field(
         alias="roleName",
         min_length=1,
diff --git a/src/skills/suggestion.py b/src/skills/suggestion.py
index c02fab2..6fd14ad 100644
--- a/src/skills/suggestion.py
+++ b/src/skills/suggestion.py
@@ -137,7 +137,7 @@ def suggest_skills(
     llm: LLMClient,
     store: VectorStore,
     *,
-    project_id: str,
+    project_id: str | None = None,
     role_name: str,
     role_description: str = "",
     project_industry: str | None = None,
@@ -164,7 +164,7 @@ def suggest_skills(
     cache = bm25_cache if bm25_cache is not None else get_bm25_cache()
 
     chunks: list[ScoredChunk] = []
-    if store.count() > 0:
+    if project_id is not None and store.count() > 0:
         chunks = hybrid_retrieve(
             question=query,
             llm=llm,
diff --git a/tests/api/test_skills.py b/tests/api/test_skills.py
index 8f4527c..396db6b 100644
--- a/tests/api/test_skills.py
+++ b/tests/api/test_skills.py
@@ -141,11 +141,46 @@ def test_suggest_skills_api_degradation_on_parse_error(
     assert response.json() == {"suggestions": []}
 
 
+def test_suggest_skills_api_200_without_project_id(
+    client: tuple[TestClient, StubLLMClient, StubVectorStore],
+) -> None:
+    http, _, _ = client
+
+    # projectId omitted from request
+    payload = {
+        "roleName": "Frontend Developer",
+        "roleDescription": "Builds user interfaces",
+        "projectIndustry": "Healthcare",
+        "availableSkills": [
+            {
+                "id": "s1",
+                "name": "React",
+                "category": "Frontend & UI",
+                "universal": False,
+            },
+            {
+                "id": "s2",
+                "name": "Communication",
+                "category": "Soft Skills",
+                "universal": True,
+            },
+        ],
+    }
+
+    response = http.post(f"{_BASE}/suggest", json=payload)
+    assert response.status_code == 200, response.text
+    data = response.json()
+    assert "suggestions" in data
+    # Only universal skill survives grounding without project context
+    assert len(data["suggestions"]) == 1
+    assert data["suggestions"][0]["name"] == "Communication"
+
+
 def test_suggest_skills_api_422_missing_fields(
     client: tuple[TestClient, StubLLMClient, StubVectorStore],
 ) -> None:
     http, _, _ = client
 
-    # Missing roleName and projectId
+    # Missing required roleName
     response = http.post(f"{_BASE}/suggest", json={})
     assert response.status_code == 422
diff --git a/tests/skills/test_suggestion.py b/tests/skills/test_suggestion.py
index 13c2a48..7cfff63 100644
--- a/tests/skills/test_suggestion.py
+++ b/tests/skills/test_suggestion.py
@@ -5,7 +5,7 @@
 from api.schemas import SkillCatalogItem
 from llm.base import Message
 from llm.errors import LLMUnavailableError
-from rag.types import Chunk
+from rag.types import Chunk, ScoredChunk
 from skills.suggestion import _build_prompt, suggest_skills
 from tests.stubs.llm import StubLLMClient
 from tests.stubs.store import StubVectorStore
@@ -386,3 +386,110 @@ def test_build_prompt_structure() -> None:
     assert "Role: Fullstack Developer" in messages[1]["content"]
     assert "Description: Works on backend and frontend" in messages[1]["content"]
     assert "Project industry/domain: Fintech" in messages[1]["content"]
+
+
+def test_suggest_skills_without_project_id_skips_retrieval() -> None:
+    store = _make_store_with_chunks()
+    catalog = [
+        SkillCatalogItem(
+            id="s1", name="React", category="Frontend & UI", universal=False
+        ),
+        SkillCatalogItem(
+            id="s2", name="Communication", category="Soft Skills", universal=True
+        ),
+    ]
+
+    llm_payload = {
+        "suggestions": [
+            {
+                "name": "React",
+                "category": "Frontend & UI",
+                "reason": "React components found",
+                "confidence": "high",
+                "universal": False,
+                "isNew": False,
+                "chunkIds": ["c-react"],
+            },
+            {
+                "name": "Communication",
+                "category": "Soft Skills",
+                "reason": "Essential for teamwork",
+                "confidence": "high",
+                "universal": True,
+                "isNew": False,
+                "chunkIds": [],
+            },
+        ]
+    }
+    llm = StubLLMClient(generate_response=json.dumps(llm_payload))
+    llm.embedding = _EMBED
+
+    # project_id is omitted / None
+    result = suggest_skills(
+        llm=llm,
+        store=store,
+        project_id=None,
+        role_name="Developer",
+        project_industry="Banking",
+        available_skills=catalog,
+    )
+
+    # React is dropped because no retrieval occurred; Communication passes
+    assert len(result.suggestions) == 1
+    assert result.suggestions[0].name == "Communication"
+    assert result.suggestions[0].chunk_ids == []
+
+
+def test_suggest_skills_without_project_id_does_not_query_store() -> None:
+    class QueryCountingStore(StubVectorStore):
+        query_count = 0
+
+        def query(self, *args: object, **kwargs: object) -> list[ScoredChunk]:
+            self.query_count += 1
+            return super().query(*args, **kwargs)
+
+    store = QueryCountingStore()
+    store.add(
+        [
+            Chunk(
+                id="c1",
+                artifact_id="a1",
+                filename="README.md",
+                text="project info",
+                embedding=_EMBED,
+                project_ids=("proj-1",),
+            )
+        ]
+    )
+    catalog = [
+        SkillCatalogItem(
+            id="s1", name="Teamwork", category="Soft Skills", universal=True
+        )
+    ]
+    llm_payload = {
+        "suggestions": [
+            {
+                "name": "Teamwork",
+                "category": "Soft Skills",
+                "reason": "Team collaboration",
+                "confidence": "high",
+                "universal": True,
+                "isNew": False,
+                "chunkIds": [],
+            }
+        ]
+    }
+    llm = StubLLMClient(generate_response=json.dumps(llm_payload))
+    llm.embedding = _EMBED
+
+    result = suggest_skills(
+        llm=llm,
+        store=store,
+        project_id=None,
+        role_name="Developer",
+        available_skills=catalog,
+    )
+
+    assert store.query_count == 0
+    assert len(result.suggestions) == 1
+    assert result.suggestions[0].name == "Teamwork"