8 failure modes documented from real incidents operating AI agent fleets.
| ID | Pattern | Severity | Category |
|---|---|---|---|
| BP-001 | Inference Over Execution | P1 | Input access / provenance |
| BP-002 | False Blocker Reporting | P2 | False status reporting |
| BP-003 | Governance Phase Skip | P2 | Governance violation |
| BP-004 | Scope Creep | P2 | Scope violation |
| BP-005 | Completion Without Verification | P2 | Quality assurance failure |
| BP-006 | Work Order Contamination | P1 | Cross-contamination |
| BP-007 | Selective Reporting | P3 | Information integrity |
| BP-008 | Authority Assumption | P3 | Role boundary violation |