fix: validate create-thread evidence boundaries

Author: Howe829

docs/superpowers/specs/2026-07-25-create-thread-capability-evidence-design.md

@@ -1,7 +1,7 @@
 # TFD create_thread Capability Evidence Design
 
-**Status:** Approved design for implementation planning  
-**Date:** 2026-07-25  
+**Status:** Approved design for implementation planning
+**Date:** 2026-07-25
 **Scope:** Add current-session `create_thread` tool-schema evidence to the live model scanner without weakening the existing account-catalog or `spawn_agent` evidence boundaries.
 
 ## 1. Problem
@@ -18,7 +18,7 @@ TFD 2.1.0 correctly scans the active Codex account model catalog and accepts the
 }
 ```
 
-Live acceptance on 2026-07-25 proved that the current `create_thread` schema advertises and successfully executes all seven visible account models. The scanner cannot represent that evidence, so its machine-readable backend result diverges from the Coordinator's live tool evidence.
+Live acceptance on 2026-07-25 proved that current `create_thread` tool metadata advertises seven visible account-model identifiers and that supplying those identifiers produces a seven-model account/backend metadata intersection. It did not prove seven completed invocations. The scanner cannot represent that evidence, so its machine-readable backend result diverges from the Coordinator's live tool evidence.
 
 ## 2. Goals
 

plugins/tfd/scripts/model_catalog.py

@@ -333,6 +333,7 @@ def normalize_catalog(
     runtime_allowed: tuple[str, ...],
     create_thread_allowed: tuple[str, ...] | None = None,
 ) -> dict[str, object]:
+    create_thread_allowed = parse_create_thread_allowed(create_thread_allowed)
     by_id: dict[str, dict[str, object]] = {}
     field_map = {
         "displayName": "display_name",
@@ -503,6 +504,7 @@ def scan_catalog(
     runtime_allowed: tuple[str, ...],
     create_thread_allowed: tuple[str, ...] | None = None,
 ) -> dict[str, object]:
+    create_thread_allowed = parse_create_thread_allowed(create_thread_allowed)
     deadline = time.monotonic() + DEFAULT_TIMEOUT_SECONDS
     with AppServerSession(codex_bin, deadline) as session:
         session.initialize()

plugins/tfd/scripts/test_model_catalog.py

@@ -366,6 +366,21 @@ def test_scanner_returns_a_sanitized_blocker_when_the_server_cannot_start(self)
 
         self.assertEqual(raised.exception.code, "launch_failure")
 
+    def test_scan_rejects_blank_create_thread_allowlist_before_launch(self) -> None:
+        with tempfile.TemporaryDirectory() as directory:
+            missing_bin = str(Path(directory) / "missing-codex")
+            with self.assertRaises(model_catalog.CatalogBlocker) as raised:
+                model_catalog.scan_catalog(
+                    missing_bin,
+                    ("model-spawn",),
+                    (" ",),
+                )
+
+        self.assertEqual(
+            raised.exception.code,
+            "invalid_create_thread_allowlist",
+        )
+
     def test_post_launch_selector_setup_failure_reaps_the_child(self) -> None:
         stdin = mock.Mock()
         stdin.closed = False
@@ -936,6 +951,19 @@ def test_scan_deadline_covers_notification_stream(self) -> None:
 
 
 class ModelCatalogNormalizationTests(unittest.TestCase):
+    def test_normalize_catalog_rejects_blank_create_thread_allowlist(self) -> None:
+        with self.assertRaises(model_catalog.CatalogBlocker) as raised:
+            model_catalog.normalize_catalog(
+                ({"id": "model-account", "hidden": False},),
+                ("model-account",),
+                (" ",),
+            )
+
+        self.assertEqual(
+            raised.exception.code,
+            "invalid_create_thread_allowlist",
+        )
+
     def test_create_thread_backend_is_available_and_preferred(self) -> None:
         payload = model_catalog.normalize_catalog(
             (
@@ -984,6 +1012,30 @@ def test_create_thread_backend_is_available_and_preferred(self) -> None:
             "UNSUPPORTED",
         )
 
+    def test_non_overlapping_create_thread_models_fall_back_to_spawn_agent(
+        self,
+    ) -> None:
+        payload = model_catalog.normalize_catalog(
+            ({"id": "model-spawn", "hidden": False},),
+            ("model-spawn",),
+            (" model-backend-only ",),
+        )
+
+        self.assertEqual(
+            payload["backend_capabilities"]["create_thread"]["models"],
+            ["model-backend-only"],
+        )
+        self.assertEqual(payload["create_thread_eligible_models"], [])
+        self.assertEqual(
+            payload["execution_readiness"],
+            {
+                "status": "PASS",
+                "code": "verified_execution_backend",
+                "backend": "spawn_agent",
+                "models": ["model-spawn"],
+            },
+        )
+
     def test_recommendation_catalog_contains_every_visible_model_with_evidence(
         self,
     ) -> None:

plugins/tfd/scripts/test_team_skill_contract.py

@@ -90,6 +90,10 @@ def test_team_requires_fresh_live_model_evidence_before_roster_proposal(
         self.assertIn("<plugin-root>/scripts/model_catalog.py", self.content)
         self.assertIn("--create-thread-model", self.content)
         self.assertIn("--runtime-allowed-model", self.content)
+        self.assertIn(
+            "pass each exact create_thread model identifier once with its own repeated `--create-thread-model` flag",
+            self.normalized,
+        )
         self.assertIn("omit `--create-thread-model`", self.content)
         self.assertIn("must not pass cached identifiers", self.normalized)
         self.assertIn(

plugins/tfd/skills/team/SKILL.md

@@ -52,7 +52,7 @@ Choose no fixed roster size. Include a role only when it adds a required special
 
 ## Verify live model evidence
 
-Before proposing a roster, the Coordinator must read the current create_thread model allowlist from live tool metadata and the current spawn_agent model allowlist from live tool metadata. Pass every exact current-session identifier to `<plugin-root>/scripts/model_catalog.py` once for its backend. For spawn_agent, pass every exact model name to `<plugin-root>/scripts/model_catalog.py` as a repeated `--runtime-allowed-model` argument. Use this pattern:
+Before proposing a roster, the Coordinator must read the current create_thread model allowlist from live tool metadata and the current spawn_agent model allowlist from live tool metadata. Pass each exact create_thread model identifier once with its own repeated `--create-thread-model` flag. For spawn_agent, pass every exact model name to `<plugin-root>/scripts/model_catalog.py` as a repeated `--runtime-allowed-model` argument. Use this pattern:
 
 ```bash
 python3 <plugin-root>/scripts/model_catalog.py \