feat: /debate skill — structured multi-model debate with convergence detection#736
Open
stedfn wants to merge 3 commits intogarrytan:mainfrom
Open
feat: /debate skill — structured multi-model debate with convergence detection#736stedfn wants to merge 3 commits intogarrytan:mainfrom
stedfn wants to merge 3 commits intogarrytan:mainfrom
Conversation
…detection Two AI models (Claude + Codex) argue a specific issue in structured rounds. Each round: position, evidence with file:line citations, rebuttal, concession, recommended resolution. Stops on convergence, concession, or round cap. Supports --interactive/-i for mid-debate checkpoints. Falls back to Claude adversarial subagent when Codex is unavailable.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
stedfn
changed the title
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
/debatepits Claude against Codex (or a Claude adversarial subagent when Codex isn't installed) in structured rounds. Each round: position, evidence with file:line citations, rebuttal, concession, recommended resolution. Debate stops when both sides converge, one concedes, or the round cap hits. User is always the final judge.--interactive/-ifor opt-in mid-debate checkpoints where you can bail, redirect, or continue after seeing each round's arguments. Without the flag, rounds run automatically.CodexvsClaude (adversarial fallback))./codex review. We tested lower reasoning effort and it produced shallower arguments that missed real issues. The cost lever is--max-rounds(default 3, hard cap 5), not reasoning quality. Optimizations can be added in follow-ups.Test plan
bun testpasses (skill validation picks up newdebate/directory)test/skill-e2e-debate.test.tscovers core debate flow and convergence detection/debate should we use Redis or Postgres for session storageproduces structured rounds with file:line evidence/debate -i <topic>pauses after round 1 with AskUserQuestion🤖 Generated with Claude Code