Create FAILURE-PROTOCOL.md

Create FAILURE-PROTOCOL.md documenting standard operating procedures for handling agent failures, stuck tasks, pipeline errors, and recovery steps. Include escalation paths and rollback procedures.

Acceptance Criteria#

1. Agent behavioral failures documented (hallucination, loops, stubborn wrongness, silent failure)
2. Pipeline & process failures documented (QC gate, intake gate, execution gate, review gate, invalid status transitions, relay failures)
3. Stuck task SOP with concrete investigation steps
4. Rollback procedures for: task-level, config, git, deployment, and pipeline state
5. Recovery verification checklist
6. Quick reference cards for each section
7. Cross-referenced with existing FAILURE-HANDLING.md and FAILURE-HANDLING-PROTOCOL.md to avoid overlap
8. INDEX.md updated to reference the new document

Notes#

• This file fills gaps in the existing failure documentation:
  • docs/FAILURE-HANDLING.md covers technical run failures (exit codes, OOM, known failure modes)
  • projects/team-management/FAILURE-HANDLING-PROTOCOL.md covers Robbie's triage/response flow
  • docs/FAILURE-PROTOCOL.md (new) covers agent behavior failures, pipeline/process failures, stuck tasks, and rollback procedures
• Rollback procedures were the biggest gap — nothing existed for task, config, git, deployment, or pipeline rollbacks before this
• The three failure docs now form a complete set covering system-level, management-level, and process-level failures