chaos

Coordinate multi-agent claim deliberation with evidence cross-examination and auditable decision receipts.

Updated May 26, 2026
One-click install
npx skills add https://github.com/XWIlluDelu/agent-share --skill chaos-xwilludelu
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: chaos
Source: https://github.com/XWIlluDelu/agent-share/tree/main/lib/chaos
Command: npx skills add https://github.com/XWIlluDelu/agent-share --skill chaos-xwilludelu

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

CHAOS helps you make high-stakes, ambiguous decisions that can’t safely rely on a single answer by forcing claim-level stress testing, dissent preservation, and evidence-standard synthesis.

Core Features & Use Cases

  • Claim-led multi-agent deliberation: the parent remains judge/final writer while advisers expose claims, evidence routes, objections, and dissent.
  • Mode selection for the right level of deliberation: supports Claim audit, Adversarial review, Council, Review loop, Deep deliberation, or Skip/direct verification.
  • Claim ledger and decision receipt: records surviving/narrowed/rejected claims and validation gaps so the final answer is auditable rather than consensus-based.
  • Evidence-standard enforcement: refuses to finalize when the required evidence standard is unmet, returning explicit failure states instead of false certainty.

Quick Start

Use CHAOS to stress-test your plan or proposal by asking: "Run CHAOS on whether we should adopt approach X for problem Y, and require evidence-backed claim validation with a minority report."

Frequently Asked Questions about chaos

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

FAQPage Schema
What is evidence-based multi-agent deliberation for ambiguous decisions?

Evidence-based multi-agent deliberation is a process where independent advisers expose claims and evidence routes for high-stakes decisions, ensuring the final synthesis is auditable rather than relying on simple consensus. It preserves dissent and rejects unsupported claims.

How do I stress-test a proposal using adversarial review?

To stress-test a proposal using adversarial review, run a mode-gated workflow that externalizes claims, cross-examines them against evidence standards, and records surviving and rejected claims in a ledger. This produces a decision receipt with explicit unresolved states.

When do I need claim-led deliberation instead of direct verification?

You need claim-led deliberation for high-value, ambiguous, or risky judgments where false certainty is dangerous. If a decision involves research planning, architecture, or proposal selection and requires surviving adversarial critique, use it instead of skipping to direct verification.

Can I get a minority report for rejected claims during research planning?

Yes, the deliberation workflow supports minority reports by preserving independent dissent and recording rejected claims in a claim ledger. This ensures that objections and validation gaps are auditable rather than discarded during the final decision synthesis.

What happens if evidence is insufficient to finalize a decision synthesis?

If evidence is insufficient to finalize a decision synthesis, the process enforces an evidence standard that refuses to finalize. It returns an explicit failure state and records the validation gaps in the decision receipt instead of manufacturing false certainty.

Are there limitations to using multi-agent deliberation for document interpretation?

A limitation of multi-agent deliberation for document interpretation is that the parent must act as judge and final writer while advisers only expose claims. The process cannot finalize decisions if the required evidence standard for the claims is unmet.