reversa-arbiter

Scores ideation options against risks and records the user's decision in decision.md.

1.5k|405|Updated Apr 26, 2026
One-click install
npx skills add https://github.com/sandeco/reversa --skill reversa-arbiter
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: reversa-arbiter
Source: https://github.com/sandeco/reversa/tree/main/agents/reversa-arbiter
Command: npx skills add https://github.com/sandeco/reversa --skill reversa-arbiter

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

When a team has brainstormed multiple solution options and identified risks, someone still has to converge on a choice with an explicit trade-off. This Skill acts as the Arbiter agent of the Reversa Ideation Team: it scores each option against fixed criteria, recommends one with the trade-off stated plainly, and records the human's decision as a traceable decision.md artifact.

Core Features & Use Cases

  • Structured scoring: Rates each option from options.md on four fixed criteria (job-to-be-done fit, effort, residual risk, legacy cost) and sums them without weighting.
  • Explicit recommendation: States which option wins, what is lost by choosing it, under what condition the recommendation changes, and what to test before committing.
  • Human decision recording: Never writes decision.md before the user explicitly chooses; registers divergences between the recommendation and the user's pick as valuable information.
  • Use Case: After running /reversa-explorer and /reversa-challenger in a legacy modernization project, invoke this agent to converge the session, produce decision.md, and advance the pipeline to the pre-spec stage.

Quick Start

Run the reversa-arbiter agent in an active ideation session to score the options, get a recommendation, and record your decision in decision.md.

Frequently Asked Questions about reversa-arbiter

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I converge a brainstorming session into a documented decision?

Run the Arbiter agent after generating framing.md, options.md, and risks.md in your ideation session. It scores each option on four fixed criteria, recommends one with the explicit trade-off, and writes decision.md only after you explicitly choose.

What criteria does the Arbiter use to score options?

It scores each option from 1 to 5 on job-to-be-done fit, effort, residual risk, and cost to the legacy system, then sums without weighting. Ties are recorded as legitimate results rather than broken artificially.

What happens if risks.md is missing from the session?

The agent stops and instructs you to run /reversa-challenger first. It requires framing.md, options.md, and risks.md from the active ideation session before it can score anything.

Can the Arbiter decide on behalf of the user?

No. It only recommends; the decision is always human. It never writes decision.md before an explicit answer, and if you choose against the recommendation, it records both your choice and the divergence without re-arguing.

Which AI agents and platforms support this Skill?

The Skill declares compatibility with Claude Code, Codex, Cursor, Gemini CLI, and other agents compatible with the Agent Skills format. It is part of the Reversa framework's Ideation Team.

What are the limitations of the Arbiter agent?

It writes only to decision.md and the current-stage field of active-ideation.json, never touches other project files, and never produces code. It also cannot proceed automatically; you must type CONTINUAR to move to the pre-spec stage.