# Debate

> Run a structured multi-AI debate through gptengage. Use when the user asks for debate or multi-model deliberation, or says /debate or $debate.

- Skill: `rahulrajaram/debate` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds@latest add rahulrajaram/debate`
- Raw SKILL.md: https://api.skillmd.com/api/skills/rahulrajaram/debate/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: rahulrajaram (https://skillmd.com/u/rahulrajaram)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/rahulrajaram/debate

---


# Debate

Read the shared
[`../../references/gptengage-invocation.md`](../../references/gptengage-invocation.md)
contract and the selected
[`../../references/gptengage-debate.md`](../../references/gptengage-debate.md)
recipe before calling a backend.

## Workflow

1. Parse the topic and debate options, validating participant selection and
   dependent flags with the operation recipe.
2. Resolve the authorized existing `gptengage` executable and run only its
   `debate` operation, passing an inner per-invocation timeout when supplied.
3. Capture the full debate, distinguish orchestration failures from individual
   backend failures, and return or integrate the requested output format.

## Boundaries

- Rounds, participant count, and synthesis increase external calls and cost;
  synthesis is an additional backend call.
- `--write` requires explicit write intent and grants no remote-write authority.
- An outer watchdog is only a secondary limit and must allow for the full
  multi-round orchestration, not merely one backend timeout.
- Never invent participants, personas, or a different topic.