Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add zgbrenner/agentcounsel --skill hallucination-red-teamgit clone --depth 1 https://github.com/zgbrenner/agentcounselWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/zgbrenner/agentcounsel/hallucination-red-team)<a href="https://agentmods.dev/skills/zgbrenner/agentcounsel/hallucination-red-team"><img src="https://agentmods.dev/badge/skills/zgbrenner/agentcounsel/hallucination-red-team/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/zgbrenner/agentcounsel/hallucination-red-team"><img src="https://agentmods.dev/badge/skills/zgbrenner/agentcounsel/hallucination-red-team.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00047 | $0.01378 |
| Opus 5 | $0.00023 | $0.00689 |
| Sonnet 5 | $0.00009 | $0.00276 |
| Haiku 4.5 | $0.00005 | $0.00138 |
Grade A, and why
Hallucination Red-Team scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 9d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 115 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Hallucination Red-Team
Purpose
Force a draft legal output to account for every factual claim, legal claim, cited authority, user-provided fact, model-generated statement, unsupported claim, and uncertainty marker. The skill produces either a safer revised version or a "not ready for use" finding.
This focused hallucination pass complements the broader Red-Team Verifier. It does not independently verify the law and does not certify that a draft is correct.
Use When
- A draft contains legal claims, factual claims, citations, quotations, or confident conclusions that may be unsupported.
- A user asks for a hallucination check, unsupported-claim audit, or safer revised version.
- A draft will be sent externally, filed, used in a client communication, or relied on for a high-risk decision.
- A model produced the draft without a clear claim/source table.
Choosing among the verification skills:
- Use this skill for a targeted hunt for fabricated or unsupported content in a draft.
- Use
red-team-verifierinstead for the full adversarial pass over an entire deliverable — the superset that also covers reasoning, framing, jurisdiction, and completeness. - Use
citation-integrity-checkinstead when the review is limited to citations and quotations. - Use
source-validationinstead for claim-by-claim classification of whether each claim in a draft is supported by the available sources.
Required Inputs
- The complete draft to red-team.
- All user-provided facts and source documents available for comparison.
- Any cited authorities, research notes, or source lists.
- The intended use and recipient.
If the draft is missing, stop and request it. If sources are missing, continue only as a limitation-heavy hallucination risk screen.
Do Not Use When
- The user wants new legal analysis instead of a verification pass.
- The output has no claims to audit.
- The user asks for a final certification that no hallucinations exist.
Legal Safety Rules
- Source and citation discipline. Follow
core/source-and-citation-discipline.md. Never invent legal authority, citations, quotations, statutes, cases, regulations, filing deadlines, or procedural rules. Label what is a provided source, a user-provided fact, an assumption, a legal inference, or an item requiring attorney verification, and use a citation placeholder such as[Attorney to insert authority]when no source is available. - Produce draft legal work product for attorney review. This is not legal advice.
- Do not claim the absence of hallucinations. State only what was checked and what remains unresolved.
- Do not invent sources, facts, citations, or safer substitute law.
- Do not assert that an authority exists or is fabricated unless verified by source material available in the session.
- Preserve privilege and confidentiality.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 9d ago First seen · 115 lines · 47 tokens per session scan A 21ca98a7aa67
Hallucination Red-Team is a skill published in the GitHub repository zgbrenner/agentcounsel (19 stars, last pushed 1mo ago), licensed MIT. It adds 47 tokens to every session and 1,378 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-03.
Other skills, from other repositories
claim-verification-checklist
Generates a point-by-point checklist of every verifiable claim in a draft article, categorized by claim type and accompanied by the specific evidence needed to confirm each one.
legal-research
A legal research skill for finding and analysing laws and court cases, including similar cases. It requires access to a reliable legal research database before producing a formal report.
document-fill
A document-filling assistant that uses a supplied template and case files to fill in forms, contracts, or legal documents. It also reports where each filled value came from and which details are missing.
contract-review
A contract-review workflow for agreements governed by Chinese law, United States law, or both. It examines clauses one by one and produces a list of legal-risk questions, with an optional marked-up Word document.
fedramp
Expert guidance for FedRAMP certification and compliance under CR26 (FedRAMP Consolidated Rules for 2026). Use this skill whenever a user asks about FedRAMP authorization, ATO (Authority to Operate), cloud security for federal government, NIST SP 800-53 controls, CSP compliance, or any of the core FedRAMP document…
eu-cra
Expert EU Cyber Resilience Act (CRA) advisor for Regulation (EU) 2024/2847 — mandatory cybersecurity and vulnerability handling requirements for all products with digital elements (PDEs) sold in the EU. Use this skill for gap analysis, product classification (Default / Class I / Class II), conformity assessment route…