| STANDALONE_EXECUTION | BLOCKING | Specialist work without PO+CA approval |
| NATO_VIOLATION | BLOCKING | Completion claims without evidence paths in tmp/ |
| COORDINATION_WITHOUT_DELEGATION | BLOCKING | PO+CA ran but raw Edit/Write used instead of specialist Task |
| SCOPE_DRIFT_BYPASS | RULES-LAYER | Prior-scope coordination logs reused for new scope |
| TEXT_OUTPUT_BYPASS | RULES-LAYER | Implementation content in text without PO+CA coordination |
| FALSE_AS_IS_STATE | RULES-LAYER | Planning on fabricated baseline not cross-validated |
| KISS_MISINTERPRETATION | RULES-LAYER | KISS applied to evidence richness instead of code complexity |
| NARROW_SEARCH_SCOPE | RULES-LAYER | Per-account search when org-wide discovery tools exist |
| JSON_ONLY_NATO | RULES-LAYER | Evidence JSON updated without updating HITL-consumable deliverable |
| AGENT_GIT_COMMIT_SUGGESTION | BLOCKING | Agent suggests or attempts git mutations — violates Principle I |
| TESTING_THEATER | BLOCKING | High pass counts but mocks without assertions, inflated omit |
| RACE_CONDITION_SCORING | RULES-LAYER | Concurrent scoring agents see different intermediate states |
| SCORING_THEATER | BLOCKING | Re-scoring unchanged artifacts — PDCA count increments, nothing changes |
| NO_ESTIMATED_COUNTS | RULES-LAYER | Numeric claims by estimation, not verification command output |
| AWS_PROFILE_SEMANTIC_MISMATCH | RULES-LAYER | Wrong profile for API — each profile has exactly ONE correct purpose |
| MOCK_CIRCULAR_VALIDATION | BLOCKING | Test asserts hardcoded mock value that code also returns — circular |
| PARALLEL_AGENT_SCOPE_BLEED | RULES-LAYER | Parallel agents create/modify files outside assigned scope |
| PLAN_DELIVERY_MISMATCH | RULES-LAYER | Sprint plan claims features not in shipped code |
| LAZY_DEFERRAL | RULES-LAYER | Claiming deferred without attempting autonomous execution first |
| HIDDEN_ASSUMPTION | RULES-LAYER | Agent picks interpretation silently without surfacing assumption |