ariffazil/Multi-Discipline Critique

Multi-Discipline Critique

Federation Eureka #4 — every non-trivial claim must survive at least 3 independent disciplines.

This is the general-purpose manual equivalent of geox_claim_challenge for claims that are not domain-specific. It hardens reasoning by forcing cross-domain self-argument before a claim is trusted, promoted, or acted upon.

Overview

This skill installs a structured cross-discipline argument protocol. The agent states a falsifiable claim, selects at least three independent disciplines, generates a falsifying challenge from each, assigns a confidence band, and synthesizes a final epistemic verdict. The goal is not to win the argument; it is to discover why the claim might be wrong before a decision commits to it.

arifOS-ACT Embedding

Before using this skill on any mutating, irreversible, or high-blast-radius task:

1. ART — Attune (what is the real task?), Recognize (what class of power?), Test (fit · authority · evidence · blast · reversible).
2. Kernel — Route to arifOS for F1–F13 judgment if action class is Maker/Messenger/Mutator/Destroyer/Sovereign.
3. ACT — Apply narrow, Constrain scope, Trace witness, STOP before corruption.
4. Receipt — Leave evidence of what changed, why, and under whose authority.

When to Use

• Before stating or relying on a non-trivial factual claim.
• Before recommending an architecture, design, or policy change.
• Before any irreversible or seal-bound decision.
• When a single domain could be producing false confidence.
• When the cost of being wrong exceeds the cost of delay.

When NOT to Use

• Do not use for trivial or definitional claims where the overhead exceeds the risk.
• Do not use as a substitute for domain-expert review when human expertise is required.
• Do not use to rationalize a predetermined conclusion; run skeptical mode if the first pass is too easy.
• Do not proceed with a claim that fails ≥2 disciplines without escalating to 888_JUDGE.

Inputs

| Input | Required | Description | |-------|----------|-------------| | claim | yes | One falsifiable sentence to be challenged | | claim_type | yes | CLAIM / PLAUSIBLE / HYPOTHESIS / ESTIMATE / UNKNOWN | | disciplines | yes | ≥3 independent challenge lenses selected for this claim | | skeptical_mode | no | If true, assume refuted and force proof otherwise (default false) |

Procedure

Step 1: State the Claim

Reduce the claim to one falsifiable sentence. Tag it:

• CLAIM — asserted as true
• PLAUSIBLE — likely but not proven
• HYPOTHESIS — candidate explanation awaiting test
• ESTIMATE — quantitative guess with error bars
• UNKNOWN — insufficient information to classify

Step 2: Select ≥3 Independent Disciplines

Choose lenses that cover different failure modes. Examples include:

| Discipline | Challenge question | |------------|--------------------| | Epistemics | Is the evidence chain intact? What is the source provenance? | | Physics / Geology | Does it violate conservation, thermodynamics, causality, or basin context? | | Law / Policy | Does it conflict with /root/SECRETS.md, F1–F13, or stated rules? | | Operations | Does it survive restart, network blip, missing env var, or degraded dependency? | | Security | Does it leak, escalate, bypass a deny rule, or widen blast radius? | | Cost / Economics | Is the resource ask proportional to the expected value? | | Ethics / Dignity | Does it reduce a person to a pattern or coerce a choice? |

Mix domains. physics + geophysics is too close; replace one.

Step 3: Challenge from Each Discipline

For each lens, answer:

1. What would falsify the claim from this perspective?
2. What evidence supports the challenge?
3. What is the confidence band? (high / medium / low / none)

Use mcp__arifos__arif_think for structured reasoning and mcp__arifos__arif_critique mode=redteam as one automatic discipline.

Step 4: Synthesize

Tally the discipline verdicts:

• All concur → promote the claim; log the witness mix.
• 1 rejects → revise the claim or downgrade its epistemic tag.
• ≥2 reject → 888 HOLD; do not use the claim as a premise.

Step 5: Output Verdict and Witness Log

Return the final tag plus a per-discipline log. If skeptical mode was enabled, note whether the claim still survived.

Allowed Tools

| Tool | Purpose | |------|---------| | Read / Bash | Inspect evidence, rules, or source material referenced by the claim | | mcp__arifos__arif_think | Structured multi-step reasoning and evidence synthesis | | mcp__arifos__arif_critique | Red-team / empathic challenge as an automatic discipline | | mcp__arifos__arif_judge | Escalate to F1–F13 judgment if the claim feeds an irreversible decision |

Forbidden Actions

• NEVER treat a claim as proven after only one discipline concurs.
• NEVER select disciplines that are structurally similar to manufacture false consensus.
• NEVER skip the discipline that seems "obvious" — that is the one that usually bites.
• NEVER hide a rejecting discipline in the final synthesis.
• NEVER promote a claim that ≥2 disciplines reject without 888_JUDGE review.
• NEVER run this skill backwards by writing the conclusion first and fitting challenges around it.

Output Format

## Skill Result: multi-discipline-critique

### Summary
One-sentence synthesis of whether the claim survives and under what epistemic tag.

### Claim
<original falsifiable claim>

### Discipline Verdicts
| Discipline | Verdict | Confidence | Falsifying Concern |
|------------|---------|------------|-------------------|
| Epistemics | concur / reject / revise | high / medium / low / none | ... |
| Security | concur / reject / revise | high / medium / low / none | ... |
| Operations | concur / reject / revise | high / medium / low / none | ... |

### Synthesis
- Promote / Revise / Hold
- Final tag: CLAIM / PLAUSIBLE / HYPOTHESIS / ESTIMATE / UNKNOWN
- Reasoning: why the claim does or does not survive

### Recommendations
- Action if the claim is promoted
- Revision if one discipline rejects
- Escalation path if ≥2 reject

### Escalations
- None / 888_JUDGE / human sovereign

Escalation Path

| Condition | Escalate To | Method | |-----------|-------------|--------| | ≥2 disciplines reject the claim | arifOS 888_JUDGE | A2A / MCP verdict_request | | Claim feeds an irreversible decision | arifOS 888_JUDGE + human | hold with reason | | Discipline selection is too narrow | Re-run with broader lens mix | internal loop | | Source evidence appears fabricated or misattributed | security.agent + arifOS judge | A2A message |

---

Skill imported from /root/.claude/skills/multi-discipline-critique.md — AAA A2A skill registry. Orthogonal tags: Δ | Governance, Evidence | HEXAGON | T1 | F2, F7, F8.