import-red-team-verifier-patrick-munro

Category: Coding Risk: Unknown ★ 3.9 · Rating 3.9/5 (8) sboghossian/mini-claude-for-legal MIT

Rating is derived from the repo's GitHub stars and shown for reference.


name: import-red-team-verifier-patrick-munro
description: Use when migrating the Patrick Munro red-team verification methodology into the mini-claude-for-legal format. This adapter maps adversarial stress-testing logic for legal documents and AI-generated legal output — deliberate challenge of assumptions, identification of counter-arguments, and verification of factual and legal accuracy — into the standard skill model. Triggers when importing any red-team or adversarial-review skill attributed to the Munro approach.
license: MIT
metadata:
id: import.red-team-verifier-patrick-munro
category: import
jurisdictions: [DIFC, ADGM, UK, UAE, multi]
priority: P3
intent: [import, red-team, verification, adversarial-review, migration]
related: [import-legal-simulation-patrick-munro, import-vendor-due-diligence-patrick-munro, import-tech-contract-negotiation-patrick-munro, import-legal-risk-assessment-anthropic]
source: Louis — HAQQ Legal AI (github.com/sboghossian/mini-claude-for-legal)
version: "1.0"

Import: Red Team Verifier (Patrick Munro)

What it does

This import adapter migrates a red-team verification skill modelled on the Patrick Munro methodology into the mini-claude-for-legal standard format. Red-teaming in the legal context means deliberately challenging the document, analysis, or argument under review — as a sophisticated adversary would — to surface weaknesses before they are exploited by a counterparty, court, or regulator.

The Munro red-team approach applies to three contexts:

  1. Document red-teaming: stress-testing a contract, legal opinion, or regulatory submission against the strongest objections
  2. AI output verification: challenging Claude-generated legal analysis to detect hallucinations, unsupported conclusions, and jurisdiction errors
  3. Argument red-teaming: testing the legal arguments in a brief or submission against the opposition's best counter-arguments

Import config

Field Source mapping Default if absent
red_team_mode Legacy mode full (document + argument + ai_output)
challenge_depth Legacy depth 3 (three levels of challenge)
adversary_persona Legacy adversary senior_opposing_counsel
hallucination_check Legacy check_hallucinations boolean true
jurisdiction_check Legacy check_jurisdiction boolean true
citation_check Legacy check_citations boolean true
output_format Legacy format red_team_report
governing_law Legacy governing_law DIFC

Dry-run preview

IMPORT PREVIEW — red-team-verifier-patrick-munro
Source shape      : Adversarial verification skill (Munro methodology)
Mode              : full (document + argument + AI output)
Challenge depth   : 3 levels
Adversary persona : senior_opposing_counsel
Hallucination check: enabled
Jurisdiction check : enabled
Citation check     : enabled
Output             : red_team_report
Governing law      : DIFC

Red-team methodology (post-import)

Level 1 — Surface challenges

  • What is the weakest factual assertion in the document?
  • What is the most ambiguous term or provision?
  • What is the most one-sided clause?
  • What statutory or regulatory provision most directly undermines the position?
  • What is the strongest precedent (or closest analogy) the opponent would cite?
  • Where does the document fail to comply with a mandatory legal requirement?

Level 3 — Strategic challenges

  • What procedural or jurisdictional argument could be used to challenge enforcement?
  • What counterparty behaviour could make this document unenforceable?
  • What change in circumstances (economic, regulatory, political) would destroy the basis for this position?

AI output verification track

When verifying AI-generated legal analysis:

  1. Citation check: verify every statute, article number, and case citation against known reliable sources. Flag any citation that cannot be verified — it may be a hallucination.
  2. Jurisdiction check: verify that every legal proposition is attributed to the correct jurisdiction and legal system (common law vs civil law; onshore vs DIFC/ADGM).
  3. Date check: verify that cited legislation is in force as of the relevant date; check for recent amendments.
  4. Internal consistency check: verify that conclusions follow from premises; flag logical gaps.
  5. Completeness check: are there mandatory requirements or standard clauses missing that a practitioner would expect?

Red-team report output

RED TEAM REPORT
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Document reviewed  : [title]
Review date        : [date]
Adversary persona  : senior_opposing_counsel
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
CHALLENGE #1 [LEVEL 1 / SURFACE]
Challenge          : [the challenge]
Weakness exploited : [specific provision or assertion]
Severity           : HIGH / MEDIUM / LOW
Recommended fix    : [specific revision]
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
[repeat for each challenge]
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
AI OUTPUT VERIFICATION
Hallucinations detected : [count] — [list]
Jurisdiction errors     : [count] — [list]
Citation warnings       : [count] — [list]
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Overall verdict : PASS / NEEDS REVISION / FAIL

MENA-specific red-team flags

  • Arabic language prevailing clause: if a bilingual contract's Arabic version differs from the English version, the Arabic governs in UAE/KSA — challenge any analysis based solely on the English text.
  • Shariah-compliance assumption: challenge any financial provision not reviewed for Shariah compliance in KSA or conservative UAE contexts.
  • Notarisation requirement: challenge enforceability of contracts in Lebanon or UAE civil jurisdictions where the transaction type requires notarial form.
  • DIFC/ADGM jurisdiction validity: challenge whether the parties actually have a legitimate connection to justify DIFC/ADGM governing law.

Failure modes

Error Likely cause Resolution
only_surface_challenges Legacy limited to Level 1 Override challenge_depth: 3
hallucination_check_disabled Legacy AI-output track disabled Enable for all AI-generated content
adversary_too_gentle Persona set to "neutral reviewer" Override to senior_opposing_counsel
jurisdiction_not_checked Citation check only Enable separate jurisdiction_check: true
  • [[import-legal-simulation-patrick-munro]]
  • [[import-vendor-due-diligence-patrick-munro]]
  • [[import-tech-contract-negotiation-patrick-munro]]
  • [[import-legal-risk-assessment-anthropic]]
  • [[review-contract-generic]]