Closed-loop work cycle that helps Claude Code problem-solve under uncertainty.
# Iliad Protocol
High-stakes debugging and problem-solving framework that prevents endless iteration cycles while maintaining quality through systematic phases and decision gates.
## Variables
PROBLEM_STATEMENT: $ARGUMENTS
ITERATION_LIMIT: 3
CONFIDENCE_THRESHOLD: 0.8
STAKES: "low" | "medium" | "high" | "critical"
COMMENTS: $ARGUMENTS
## Protocol Phases
### Phase 1: Reconnaissance (Time-boxed: 25% of budget)
```
OBJECTIVE: Understand the problem space completely
=> research_agent: Gather all available context
=> analysis_agent: Map system state and dependencies
=> time_machine: Historical context and decisions
=> knowledge_graph: Visualize relationships
GATE: Can we clearly define the problem?
- [ ] Problem statement is specific and measurable
- [ ] Root cause hypotheses are formed
- [ ] Success criteria are defined
- [ ] Risk assessment is complete
If GATE fails: Escalate or change approach
```
### Phase 2: Strategic Planning (Time-boxed: 15% of budget)
```
OBJECTIVE: Design solution approach with multiple contingencies
-> implementation_variants: Generate 3 solution approaches
-> swarm_intelligence: Evaluate approaches across dimensions
-> pattern_synthesizer: Learn from similar problems
DECISION MATRIX:
- Implementation complexity
- Risk of unintended consequences
- Resource requirements
- Probability of success
- Rollback difficulty
GATE: Do we have a viable plan?
- [ ] Primary approach selected with high confidence
- [ ] Backup approaches identified
- [ ] Risk mitigation strategies in place
- [ ] Success metrics defined
If GATE fails: Return to reconnaissance or escalate
```
### Phase 3: Controlled Implementation (Time-boxed: 45% of budget)
```
OBJECTIVE: Execute solution with continuous validation
For each iteration (max 3):
-> Implement smallest testable change
-> Validate against success criteria
-> Assess unintended consequences
-> Document findings
ITERATION_GATE:
- [ ] Progress toward objective
- [ ] No regression introduced
- [ ] Within quality thresholds
- [ ] Learning captured
If ITERATION_GATE fails:
- Try backup approach
- Reduce scope
- Escalate if at iteration limit
PHASE_GATE: Is solution working?
- [ ] Primary objectives met
- [ ] No critical regressions
- [ ] Quality maintained
- [ ] Monitoring in place
If PHASE_GATE fails: Activate contingency plan
```
### Phase 4: Validation & Documentation (Time-boxed: 15% of budget)
```
OBJECTIVE: Ensure solution is robust and knowledge is captured
-> systematic_debug: Comprehensive testing
-> evolution_tracker: Update historical context
-> pattern_synthesizer: Extract reusable patterns
-> mem0: Store decision rationale and lessons
FINAL_GATE: Is solution production-ready?
- [ ] All tests passing
- [ ] Performance verified
- [ ] Documentation updated
- [ ] Monitoring configured
- [ ] Rollback plan tested
If FINAL_GATE fails: Return to implementation or accept partial solution
```
## Decision Framework
### When to Continue vs. Stop
**Continue if:**
- Clear progress toward objectives
- No critical system damage
- Within iteration/time budget
- Learning is occurring
**Stop and Escalate if:**
- No progress after 2 iterations
- Critical system damage risk
- Problem scope expanding uncontrollably
- Stakes exceed available resources
**Accept Partial Solution if:**
- Core objective achieved (even if incomplete)
- Further iteration has diminishing returns
- Business/time constraints require it
- Sufficient foundation for future work
### Escalation Triggers
**Technical Escalation:**
- Problem requires expertise outside team
- Infrastructure/architectural changes needed
- Cross-team coordination required
**Management Escalation:**
- Resource constraints preventing solution
- Business priority conflicts
- Risk exceeds acceptable thresholds
## Quality Gates
### Code Quality
- No new linting errors
- Test coverage maintained
- Performance within bounds
- Security review passed
### System Quality
- No service degradation
- Error rates within limits
- Monitoring alerts clear
- Dependency health good
### Process Quality
- Documentation updated
- Knowledge captured in mem0
- Rollback procedures tested
- Team informed of changes
## Example Usage
```bash
/iliad-protocol "Fix MCP telemetry integration causing agent workflow completion issues"
Stakes: HIGH (affects core product functionality)
Budget: 2 days
Iteration Limit: 3
Phase 1 (Reconnaissance - 4 hours):
- Historical analysis of MCP integration
- Current system state mapping
- Problem reproduction verification
- Risk assessment
Phase 2 (Planning - 2 hours):
- Multiple fix approaches identified
- Risk vs benefit analysis
- Rollback strategy defined
Phase 3 (Implementation - 12 hours):
- Iteration 1: Minimal fix attempt
- Iteration 2: Comprehensive approach
- Iteration 3: Fallback solution
Phase 4 (Validation - 2 hours):
- Full regression testing
- Performance validation
- Documentation updates
```