citevahti-claims

citevahti-claims is a skill for Claude Code from heidihelena/citevahti. It costs 100 tokens per session (1,426 once invoked), scanned A, original, Apache-2.0.

A review guide for public writing about CiteVahti, a research-review tool. It checks website copy, documentation, release notes, social posts, slides, and captions against the product’s evidence limits.

In plain words
What is it for?
Use it before publishing or editing any public CiteVahti material, or when checking whether a specific claim about the product is supported.
Why use it?
It prevents public descriptions from promising accuracy, certification, or other results that the product does not guarantee.

Skill for Claude Code

Written for Claude Code: shipped in a Claude Code plugin.

Part of the citevahti plugin — 13 skills, 1 MCP server shipped together

Good fit Use it before publishing or editing any public CiteVahti material, or when checking whether a specific claim about the product is supported.

Compare 6 skills from other repositories ↓
Install with agentmods
npx agentmods add skills/heidihelena/citevahti/citevahti-claims
Install

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.

Any agent
npx skills add heidihelena/citevahti --skill citevahti-claims
Clone the repo
git clone --depth 1 https://github.com/heidihelena/citevahti

Made for: Claude Code.

Or install citevahti, the plugin that ships this one along with the rest of its 13 skills, 1 MCP server.

Wrote 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.

agentmods badge for citevahti-claims

README.md
[![agentmods](https://agentmods.dev/badge/skills/heidihelena/citevahti/citevahti-claims/github.svg)](https://agentmods.dev/skills/heidihelena/citevahti/citevahti-claims)
Your own site
<a href="https://agentmods.dev/skills/heidihelena/citevahti/citevahti-claims"><img src="https://agentmods.dev/badge/skills/heidihelena/citevahti/citevahti-claims/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.

agentmods 80×15 button for citevahti-claims

Your own site · 80×15
<a href="https://agentmods.dev/skills/heidihelena/citevahti/citevahti-claims"><img src="https://agentmods.dev/badge/skills/heidihelena/citevahti/citevahti-claims.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 100 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 1,426 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5.1 $0.00100 $0.01426
Opus 5 $0.00050 $0.00713
Sonnet 5 $0.00020 $0.00285
Haiku 4.5 $0.00010 $0.00143

Measured 12d ago against content hash 23c6215e0708, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-12, from the pricing page.

Security

Grade A, and why

citevahti-claims 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.

skills/citevahti-claims/SKILL.md · 84 lines

How it starts

The opening of the file, as written. The whole thing — 84 lines — stays where its author put it; the contents beside it link to each section on GitHub.

CiteVahti claims — audit public language before it ships

CiteVahti's brand core is a refusal: it never issues a verdict, and it holds itself to the evidence standards it promotes. One sentence of overclaiming — "AI-verified", "guarantees accuracy", both forbidden below — spends that credibility permanently. This skill audits every public artifact against the boundary the product actually keeps.

Canonical sources (the audit is against these, not memory): docs/DISCLOSURE.md (what is never certified) · docs/KNOWN_LIMITATIONS.md (what is honestly not done) · docs/METHODS.md §PRISMA-trAIce (framework-compliance ceiling) · docs/SAFETY_INVARIANTS.md (what "safe" concretely means).

Triggers

Use when: publishing or editing site copy, README, docs/STATUS.md positioning, LinkedIn/social posts, release notes, talk slides, marketing drafts (docs/marketing/), or answering "can we claim X?".

Do NOT use for: internal notes and ADRs (not public), or code review.

The must-not-claim list

Reject or rewrite any artifact that states or implies:

Forbidden Why Say instead
"guarantees accuracy / correctness" no accuracy promise can be made; recall < 100% is a documented product fact "surfaces problems for human adjudication"
"AI-verified" / "verified by AI" AI is a blinded advisory second rater; it never sets the final value "human-decided, AI-assisted, blinded second opinion"
"catches all errors" / "no citation problem escapes" detection is imperfect and DOI/PMID-keyed; untestable claims exist "flags what it can check; marks the rest untestable"
certification language: "certifies", "approves", "CiteVahti-validated", badges/seals DISCLOSURE.md: use does not certify truth, quality, or absence of problems "structured decision support and an audit trail"
"PRISMA / PRISMA-trAIce / Cochrane compliant, endorsed, aligned" METHODS.md explicitly disclaims compliance and endorsement "mirrors dual-screening logic; PRISMA-trAIce / RAISE-style transparency reporting"
verdict language: "CiteVahti says this citation is wrong" it records your judgment with provenance "flagged for your review", "did not find support"
accuracy numbers with no source until the eval ledger is scored, there are no numbers (KNOWN_LIMITATIONS.md) cite the published eval page, or phrase as commitment ("we will publish…")
"your data never leaves your machine" (absolute) literature lookups go to PubMed/OpenAlex/Semantic Scholar/Crossref "manuscript and ratings stay local; only literature queries go out; no telemetry"
"catches hallucinated references" (absolute) fabricated-reference detection keys on DOI/PMID; identifier-less items can't be checked "refuses to write unverifiable citations; flags references that don't resolve"
hosted-service implications: "upload", "try it in your browser" for the panel ADR-0007: the web app is a local loopback panel; nothing is hosted "runs locally; ten-minute install"

Read the full file on GitHub · 84 lines

Changes

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.

  1. 12d ago First seen · 84 lines · 100 tokens per session scan A 23c6215e0708

Subscribe to this mod's changes

citevahti-claims is a skill published in the GitHub repository heidihelena/citevahti (1 stars, last pushed 7d ago), licensed Apache-2.0. It adds 100 tokens to every session and 1,426 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.