fix(skills_guard): explain why --force is rejected on dangerous verdicts

Follow-up to @sprmn24's verdict-logic fix. The previous block-message
ended in 'Use --force to override' regardless of verdict — but as of
the --force fix above, dangerous community/trusted skills can't be
overridden by --force at all. The misleading hint sends users in a
loop. Replace it with a specific message that tells them what the
documented behavior actually is.

Adds two regression tests covering the dangerous-verdict message
shape and one that pins the existing --force hint for non-dangerous
blocks.

tests/tools/test_skills_guard.py

@@ -152,6 +152,31 @@ class TestShouldAllowInstall:
         )
         assert allowed is False
         assert "Blocked" in reason
+        # Error message MUST explain why --force didn't work, not invite a retry.
+        assert "does not override" in reason
+        assert "Use --force to override" not in reason
+
+    def test_force_does_not_override_dangerous_for_trusted_message(self):
+        f = [Finding("x", "critical", "c", "f", 1, "m", "d")]
+        allowed, reason = should_allow_install(
+            self._result("trusted", "dangerous", f), force=True
+        )
+        assert allowed is False
+        assert "does not override" in reason
+        assert "Use --force to override" not in reason
+
+    def test_non_dangerous_block_keeps_force_hint(self):
+        # When --force CAN override the block, the error message must still
+        # point to it. Use builtin trust + dangerous to land in the block
+        # branch without triggering the dangerous-specific message.
+        f = [Finding("x", "high", "network", "f", 1, "m", "d")]
+        # Construct a path where decision == block but verdict != dangerous.
+        # community + caution = block per current INSTALL_POLICY.
+        allowed, reason = should_allow_install(
+            self._result("community", "caution", f), force=False
+        )
+        assert allowed is False
+        assert "Use --force to override" in reason
 
     def test_force_does_not_override_dangerous_for_trusted(self):
         f = [Finding("x", "critical", "c", "f", 1, "m", "d")]

tools/skills_guard.py

@@ -674,6 +674,13 @@ def should_allow_install(result: ScanResult, force: bool = False) -> Tuple[bool,
             f"{len(result.findings)} findings)"
         )
 
+    # Dangerous verdicts cannot be overridden by --force (community/trusted);
+    # other blocks can.
+    if result.verdict == "dangerous" and result.trust_level in ("community", "trusted"):
+        return False, (
+            f"Blocked ({result.trust_level} source + dangerous verdict, "
+            f"{len(result.findings)} findings). --force does not override a dangerous verdict."
+        )
     return False, (
         f"Blocked ({result.trust_level} source + {result.verdict} verdict, "
         f"{len(result.findings)} findings). Use --force to override."