dabydat/alpha-omega/plan

neuron: plan

• triggers_when: the task is non-trivial (>1 file, architectural or >2 hours)
• tier: T3
• budget: 1200 tokens
• eval: plan with phases, dependencies and acceptance criteria -> threshold 1
• reward: pheromone +1 if eval >= threshold

Procedure

1. Read memory/STATE.md and existing diagrams (without writing code).
2. Divide into phases with agent, files, dependencies and acceptance criteria.
3. Identify parallel lanes and risks with mitigation.
4. Estimate complexity (Low/Medium/High) and token budget.
5. Get approval and launch /implement.

Checklist

• [ ] Phases with files and assigned agent
• [ ] Dependencies and parallel lanes identified
• [ ] Testable acceptance criteria
• [ ] Risks with mitigation

Pitfalls

• Planning trivial tasks (<1 file)
• Writing code during planning
• Omitting dependencies between phases