Paseo Committee

When you're stuck in a loop or tunnel vision has set in, this spins up two high-reasoning agents from different providers (Claude Opus with extended thinking and GPT-5.4 by default) to do root cause analysis in parallel. They never touch code, just think. You feed them the problem, challenge their assumptions, synthesize a plan from their responses, implement it yourself or via a delegate, then send the diff back for review. The committee stays alive through the whole cycle so they can catch drift. It's designed to force you to step back rather than double down, and the hard rule is no edits from committee members, analysis only.

skills.sh installs: 5.2K
GitHub stars: 23.7K
last updated: Aug 21, 2026

View on GitHub

Install

$ npx -y skills add getpaseo/paseo --skill paseo-committee --agent claude-code

Installs into .claude/skills of the current project.

Files

Committee Skill

Two agents from contrasting profiles, fresh context, planning a solution in parallel.

Use the user's request and current conversation as input, including any arguments supplied with the skill invocation.

Prerequisites

Read the paseo skill. Call list_profiles before choosing committee members. Do not create committee agents until you have read the configured profiles and their notes.

Contrast is the point of a committee, so pick profiles from different provider families when possible. Materialize each profile into create_agent.

Composition

Two members with different reasoning styles, selected from configured Agent profiles:

one whose notes fit planning, research, or root-cause analysis
one contrasting high-reasoning profile from another provider family

If the user names profiles, use those. If fewer than two suitable profiles are configured, use Paseo's provider discovery fallback for the missing member and tell the user. Override the selection only when the user explicitly asks for different members.