| name | debate-decisions |
| description | A compact decision protocol for non-trivial choices, reviews, governance calls, and live disagreement. Use implicitly at decision points, even when nobody says "debate", whenever a choice is high-impact, hard to reverse, policy-bearing, or has multiple plausible paths that evidence alone cannot settle. Skip for trivial, cheap-to-reverse, or deterministically decidable choices. |
Debate Protocols
Use debate as a small decision aid, not a format museum. If grep, tests, logs, source files, or official records can decide the question, use them as the judge and skip debate.
Core Rules
- State the resolution in one sentence.
- Assign sides before arguing. For peer agents, the proposer argues against their own proposal and the reviewer argues for it; otherwise use a coin flip or a stable hash of the resolution.
- Treat assigned side as a role, not belief. Concession is allowed and counts as progress.
- The proposer must argue against their own proposal at least once.
- End with a decision record: winner, strongest argument, unresolved risk, evidence that would flip the result, and preserved dissent.
- Make every dissent testable. An objection must name the evidence that would prove or disprove it, or be labeled speculation.
- Cap the loop: one rebuttal cycle by default. If positions do not change after two exchanges, stop and record the unresolved dissent.
Compressed Rounds
Use compressed rounds by default inside normal work:
Resolution:
Pro constructive:
Con constructive:
Crossfire:
Rebuttals:
Weighing:
Decision:
Preserved dissent:
For multi-agent HCOM work, send each role its assignment and require a short constructive plus one attack on its own side.
Read references/termination-and-concession.md when a debate risks looping, when agents disagree after one rebuttal cycle, or when dissent starts becoming performative.
Output Shape
Decision:
Why:
Best opposing argument:
What would flip it:
Next action:
