| name | 1-3-1 |
| description | Use when a problem needs fast candidate generation and narrowing: run 3 independent fix proposals, compare by constraints/risk/complexity, then execute one selected path. |
1-3-1 strategy
Goal
Resolve a problem by first understanding it deeply, then collecting multiple candidate fixes, then converging on one decision with the main agent.
Process
-
Identify the problem well
- Restate the goal and constraints in one sentence.
- List what is known, unknown, and assumed.
- Define success criteria and hard constraints.
- Call out tradeoffs and potential risks before proposing changes.
-
Fire up 3 subagents
- Launch 3 independent subagents on the same problem statement.
- Each subagent must return:
- a candidate fix
- rationale
- risks
- expected validation checks
- Subagents should prefer concrete, minimal, low-complexity approaches.
-
Main agent selects one fix
- Compare the three proposals against the success criteria.
- Prefer the option with the best risk/reward balance and least complexity.
- Merge useful parts only if conflicts are resolved explicitly.
- Choose one final path and execute it.
Scoring rubric for selection
- Correctness against stated constraints
- Simplicity and maintainability
- Failure-risk and rollback ease
- Execution speed
Anti-patterns to avoid
- Skipping problem framing and jumping to implementation
- Letting the loudest proposal override evidence
- Combining all proposals without adjudication
Exit condition
- The main agent has one chosen implementation plan and a short justification tied to the rubric.
