Summary

Document the trust-based compliance antipattern in architecture governance to prevent future protocol design errors.

Background

From PR co-mingling retrospective (PR #669): Trust-based compliance ("agent should remember to check branch") failed, while verification-based enforcement ("tool output MUST appear in transcript") succeeds.

Priority: P1 (Prevents future design errors)

Specification

Skill ID: protocol-014
Atomicity: 94%
Category: Architecture governance

Antipattern Definition

Trust-Based Compliance: Protocol requirements that rely on agent memory/behavior without verification.

Failure Mode: Agents forget, context limits cause omission, instructions drift over time.

Success Rate: ~0% for non-trivial protocols

Replacement Pattern

Verification-Based Enforcement: Protocol requirements that generate observable artifacts that can be verified.

Success Indicators: Tool output in transcript, files on disk, API responses, git history.

Success Rate: 90%+ when properly designed

Documentation Location

Create or update: .agents/governance/PROTOCOL-ANTIPATTERNS.md

Proposed Content Structure

1. Antipattern: Trust-Based Compliance
   • Description, examples, why it fails
   • Evidence from PR docs(retrospective): PR co-mingling root cause analysis #669
2. Pattern: Verification-Based Enforcement
   • Description, examples, why it works
   • Design checklist
3. Case Studies
   • Branch verification (PR docs(retrospective): PR co-mingling root cause analysis #669)
   • Test execution
4. Design Guidelines
   • Observable, verifiable, blocking criteria

Implementation

1. Create .agents/governance/PROTOCOL-ANTIPATTERNS.md
2. Document trust-based compliance antipattern
3. Document verification-based enforcement pattern
4. Add case studies (branch verification, test execution)
5. Link from SESSION-PROTOCOL.md and AGENT-INSTRUCTIONS.md

Testing

• Review existing protocols for trust-based patterns
• Identify candidates for verification-based redesign
• Validate case studies with retrospective data

Related

• PR docs(retrospective): PR co-mingling root cause analysis #669: Root cause analysis (primary evidence)
• Issue feat(protocol): add mandatory branch verification to SESSION-PROTOCOL #684: Branch verification (case study)
• Skill: protocol-014 (trust-based compliance antipattern)

Acceptance Criteria

• PROTOCOL-ANTIPATTERNS.md created
• Trust-based compliance antipattern documented with evidence
• Verification-based enforcement pattern documented
• At least 2 case studies included
• Design checklist provided
• Linked from SESSION-PROTOCOL.md and AGENT-INSTRUCTIONS.md

Impact

Prevents future protocol design errors by documenting failure patterns and providing replacement patterns with evidence.