Borrowing it
Nothing to install: this file belongs to heidihelena/vahtian. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/heidihelena/vahtian/main/.claude/skills/vahtian-brand-safety/SKILL.mdgit clone --depth 1 https://github.com/heidihelena/vahtianWrote 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/heidihelena/vahtian/vahtian-brand-safety)<a href="https://agentmods.dev/skills/heidihelena/vahtian/vahtian-brand-safety"><img src="https://agentmods.dev/badge/skills/heidihelena/vahtian/vahtian-brand-safety/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/heidihelena/vahtian/vahtian-brand-safety"><img src="https://agentmods.dev/badge/skills/heidihelena/vahtian/vahtian-brand-safety.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 | $0.00108 | $0.01705 |
| Opus 5 | $0.00054 | $0.00852 |
| Sonnet 5 | $0.00022 | $0.00341 |
| Haiku 4.5 | $0.00011 | $0.00170 |
Grade A, and why
vahtian-brand-safety 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 12d 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 — 100 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Vahtian brand-safety reviewer
This skill keeps the site from drifting into AI hype. Vahtian is built by a clinician (MD, PhD); overclaiming is both a trust risk and a compliance risk.
This is the single canonical home of the invariant and copy rules. The
other skills (vahtian-ux-auditor, vahtian-frontend-implementer, CLAUDE.md)
defer here rather than restating them. In-repo source of truth for marketing
claims: AD_CLAIMS.md — on any conflict between this skill and AD_CLAIMS.md,
AD_CLAIMS.md wins; update this skill to match it, never the reverse.
The invariant (never let copy break this)
Vahtian assesses whether a cited source supports a specific claim, and records who decided what — with a human first, AI second, and an auditable trail. It does NOT certify scientific truth, clinical validity, manuscript quality, publication readiness, or the absence of citation problems.
Three rules that follow from it:
- The human decides. AI is a blinded second opinion the human can ignore, offered only after the human has rated. Copy must never make AI the judge.
- Support, not truth. Vahtian checks whether the cited source supports the sentence — not whether the sentence is true.
- No unbenchmarked accuracy claims. No percentages, no "catches every…", no "reviewer-proof" — there is no published validation study.
Forbidden / risky phrasings — flag every one
| ❌ Don't say | Why it's unsafe |
|---|---|
| "verifies truth", "fact-check your paper", "make sure your claims are correct" | Checks support, not truth |
| "guarantees", "proves", "100% safe", "never miss a bad citation" | Guarantee with no benchmark |
| "fully automated review", "AI decides", "AI checks your citations" | Makes AI the judge — breaks human-first |
| "AI-drafted/AI-generated peer review", "let AI review the manuscript", "outsource the review to AI" | A peer reviewer accepts confidentiality, accountability, and IP-protection duties; drafting a review with a third-party AI breaks all three at once (Brem, Chiligireddy & Worthington, IEEE Eng. Manag. Rev. 2026, 10.1109/EMR.2026.3702480). Review-side tools assist a human on narrow, local, verifiable checks; they never draft, generate, or recommend the review, and nothing uploads |
| "publication-ready", "reviewer-proof" | Certifies publication suitability |
| "eliminates bias" | Overclaim; Vahtian records and preserves disagreement |
| "detects all citation errors", "99% accurate" | Accuracy claim without published study |
| "certifies evidence", "verify your science", "clinical validity" | Certification / truth framing |
| "anonymous" (for contributed data) | Always "de-identified"; contribution is opt-in, default-off |
| "future-proof", "will open in any editor in twenty years", "works forever" | Guarantee-shaped promise about longevity — say "most likely to still open", "opens in any editor today" |
| "studies show…", "most researchers find…" with no source | Evidence claim without provenance — cite it or rewrite as reasoning from mechanism |
| "NVivo/MAXQDA cannot do X" (or any competitor claim) unchecked | False competitor claims are the fastest trust kill with this audience — check the competitor's own docs; sell the differentiator that survives |
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.
- 12d ago First seen · 100 lines · 108 tokens per session scan A 35c5671691a7
vahtian-brand-safety is a skill published in the GitHub repository heidihelena/vahtian (1 stars, last pushed 1mo ago), licensed Apache-2.0. It adds 108 tokens to every session and 1,705 once invoked, about $0.0005 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 2026-08-31.
Other skills, from other repositories
link-evidence
Prevents fabricated URLs. Only use links that appear in the sources pool or are provided by the user.
Web Research Report
Synthesize fetched web sources into a structured research report on a topic: key findings, themes, tradeoffs, and a recommendation, with citations. Use for web research, literature review, and report writing.
paper-glossary
Use when building reusable Obsidian glossary notes from an existing paper source manifest, optionally with a raw-sections override, especially when a reader needs a reviewed shortlist before glossary notes and article links are changed.
smart-sourcing
Selects optimal sources for tool calls, balancing accuracy with token cost. Use before research tasks or when deciding whether a claim needs verification.
docx-press
Teaches agents how to author Markdown that renders cleanly into DOCX via writedocx, with heading hierarchy (up to 6 levels), tables, lists, blockquotes, code blocks, citations, embedded images, SVG diagrams (rendered to PNG first), and document structure patterns for reports, proposals, and technical documents.
3d-deep-research
A research workflow for investigating products, companies, technologies, people, markets, industries, and complex events. It produces a traceable report based on sources, evidence, and analysis of how events developed, what forces shaped them, and how the parts work.