What problem does it solve? Debugging complex bugs often suffers from single-pass reasoning errors and confirmation bias. This Skill orchestrates a structured multi-agent debate where N independent solvers argue over R rounds and an isolated judge with a frozen rubric produces one synthesized recommendation for a registered bug. ## Core Features & Use Cases - Three Debate Modes: diagnosis (competing causal hypotheses), repair (competing fix strategies), and spec (recommendation on whether code, tests, or spec is correct). - Isolated Judge with Frozen Rubric: The judge sees only anonymized, shuffled final proposals, preventing echo chambers and reward hacking. - Auditable State on Disk: All rounds, convergence metrics, and the final answer are written to the bug's debate folder; failures are recorded with explicit status and never silently replaced. - Optional External Harnesses: Detects installed CLI agents (Codex, Gemini CLI, opencode) and can include them as solvers or critics only with explicit user consent; restricted-visibility bugs never leave the local folder. - Use Case: A team has a registered bug where code, tests, and spec disagree. Run the debate in spec mode with 3 solvers over 2 rounds; the judge returns a recommendation (spec-correta, spec-desatualizada, or spec-gap) with evidence, while the human keeps the final verdict. ## Quick Start Ask the agent to run a debate on bug BUG-20260715-A7K3 in diagnosis mode with 3 solvers and 2 rounds, then show the judge's recommendation.
