Комитет агентов для анализа первопричин

★ 7.7 · orchestration

paseo-committee is a Claude Code skill that assembles a two-agent committee with contrasting reasoning styles to perform root cause analysis and produce an actionable plan. The skill reads configured Paseo profiles via `list_profiles`, selects members from different provider families when possible, and launches both agents in parallel with `[Committee] <task>` titles using the same problem-level prompt. There are no polling or interruptions — models may reason for 15–30 minutes before responding. When the agents disagree, their arguments are exchanged iteratively until they converge; the final consensus, along with a summary of where they diverged, is shared with the user. A hard rule enforces analysis-only mode: every prompt ends with an explicit instruction prohibiting any file edits, creation, deletion, or code writing. Best suited for debugging loops, tunnel-vision situations, and complex planning challenges.