Ticket Workflow Skill

This skill provides the state machine and coordination logic for the complete ticket resolution workflow.

State Machine

┌─────────┐    ┌─────────┐    ┌──────────┐    ┌─────────┐
│ pending │───►│ fetched │───►│ analyzed │───►│ planned │
└─────────┘    └─────────┘    └──────────┘    └─────────┘
                    │              │               │
                    ▼              ▼               ▼
               ┌─────────────────────────────────────────┐
               │                failed                    │
               └─────────────────────────────────────────┘
                                   │
                                   │ (resume)
                                   ▼
┌─────────────────┐    ┌───────────────┐    ┌───────────┐
│ plan_validated  │───►│ implementing  │───►│ completed │
└─────────────────┘    └───────────────┘    └───────────┘

States

State	Description	Next States
pending	Workflow just started	fetched, failed
fetched	Ticket data retrieved	analyzed, failed
analyzed	Complexity determined	planned, failed
planned	Implementation plan ready	plan_validated, failed
plan_validated	Plan approved by user	implementing, failed
implementing	Auto-implementation in progress	completed, failed
completed	All phases finished	(terminal)
failed	Error occurred	(can resume)

Phase Mapping

State Transition	Phase	Skill Used
pending → fetched	fetch	fetch-ticket
fetched → analyzed	analyze	analyze-ticket
analyzed → analyzed	explore	(AEP agents)
analyzed → planned	plan	(Architect)
planned → plan_validated	validation	plan-validation
plan_validated → implementing	implement	/resolve --continue

Status File Management

Location

.claude_resolve/{ticket-id}/status.json

Initialize Status

{
  "ticket_id": "PROJ-123",
  "source": "youtrack",
  "started_at": "2024-01-15T10:30:00Z",
  "updated_at": "2024-01-15T10:30:00Z",
  "state": "pending",
  "phases": {
    "fetch": "pending",
    "analyze": "pending",
    "explore": "pending",
    "plan": "pending",
    "implement": "pending"
  },
  "options": {}
}

Update Status

After each phase completion:

{
  "state": "fetched",
  "updated_at": "2024-01-15T10:31:00Z",
  "phases": {
    "fetch": "completed",
    "analyze": "pending",
    ...
  }
}

Mark Failed

On error:

{
  "state": "failed",
  "updated_at": "2024-01-15T10:32:00Z",
  "phases": {
    "fetch": "completed",
    "analyze": "failed",
    ...
  },
  "error": {
    "phase": "analyze",
    "message": "Could not determine complexity",
    "timestamp": "2024-01-15T10:32:00Z"
  }
}

Resume Logic

Check for Existing Status

if [ -f ".claude_resolve/{ticket-id}/status.json" ]; then
    # Load and check state
fi

Resume from State

def get_resume_point(status):
    # Find last completed phase
    phase_order = ['fetch', 'analyze', 'explore', 'plan', 'implement']

    for phase in phase_order:
        if status['phases'][phase] in ['pending', 'failed']:
            return phase

    return None  # All completed

Display Resume Info

# Reprise du Workflow

**Ticket**: {ticket-id}
**Etat precedent**: {state}
**Derniere mise a jour**: {updated_at}

## Phases

| Phase | Statut |
|-------|--------|
| Fetch | completed |
| Analyze | completed |
| Explore | skipped |
| Plan | pending |
| Implement | pending |

## Point de Reprise

Reprise depuis: **plan**
Erreur precedente: {error.message}

Continuer ? [O/n]

Workflow Orchestration

Phase: Fetch

1. Check if status exists and fetch completed → skip
2. Invoke fetch-ticket skill
3. Save ticket.md
4. Update status: state = "fetched", phases.fetch = "completed"

Phase: Analyze

1. Check if status exists and analyze completed → skip
2. Invoke analyze-ticket skill
3. Determine complexity
4. Save analysis.md
5. Update status: state = "analyzed", phases.analyze = "completed", complexity = ...

Phase: Explore

1. Check complexity level
2. If SIMPLE: skip, mark phases.explore = "skipped"
3. If MEDIUM: launch 1 explore agent
4. If COMPLEX: launch 3 explore agents in parallel
5. Append findings to analysis.md
6. Update status: phases.explore = "completed"

Phase: Plan

1. If COMPLEX and not --skip-architect: invoke Architect skill
2. Generate implementation plan
3. Save plan.md
4. Update status: state = "planned", phases.plan = "completed"

Phase: Plan Validation

AUTO mode: Auto-validate, continue to implement INTERACTIVE mode: Validation loop (Valider et implémenter / Valider et arrêter / Modifier / Régénérer)

• Update status: state = "plan_validated", plan_validated_at = "{timestamp}"

Phase: Implement

AUTO mode: Always executed after plan validation INTERACTIVE mode: Via "Valider et implémenter" choice or --continue flag

1. Execute /compact to clear context
2. Delegate implementation to subagent
3. Update status: state = "implementing" → phases.implement = "completed"

Configuration Integration

Read from project's .claude/ticket-config.json:

• Merge with default config from references/default-config.json
• Override with command-line options

Priority order:

1. Command-line options
2. Project config
3. Default config

Error Recovery

Transient Errors

• Network timeout → retry with backoff
• MCP unavailable → fallback or prompt user

Persistent Errors

• Ticket not found → mark failed, clear instructions
• Permission denied → mark failed, suggest resolution

Recovery Actions

## Erreur Recuperable

L'erreur "{error.message}" peut etre resolue.

**Actions suggerees**:
1. {action 1}
2. {action 2}

**Pour reprendre**:
```bash
/resolve {ticket-id}  # Reprend automatiquement

## Files Generated

.claude_resolve/{ticket-id}/ ├── status.json # Workflow state ├── ticket.md # Original ticket content ├── analysis.md # Complexity analysis + exploration findings └── plan.md # Implementation plan

## Integration Points

### With AEP Skill
- Used for COMPLEX tickets during explore phase
- Applies ANALYSE-EXPLORE-PLAN methodology

### With Architect Skill
- Used for COMPLEX tickets during plan phase
- Ensures plan follows best practices

### With /resolve --continue
- Plans saved in standard format
- Resumes implementation from validated plan

## Language

- Status file: English (machine-readable)
- User messages: French
- Technical output: English