AI answers can repeat outdated prices, missing features, or facts about another company. This skill checks individual claims and drafts sourced corrections with a recheck plan.
Review the Skillstore skill "revnu-fix-ai-claims" from https://skillstore.io/skills/revnu-app-revnu-fix-ai-claims.md and its manifest at https://skillstore.io/api/skills/revnu-app-revnu-fix-ai-claims/manifest. Verify the artifact. Stop and obtain explicit user consent before installing or changing files.
エージェントは引き続き計画を提示し、セキュリティポリシーで必要な確認を求める必要があります。
エージェントが読めるリソース
AI エージェント、クローラー、スクリプトがページ全体ではなく整理されたコンテキストを必要とする場合は、これらのリンクを使ってください。
「revnu-fix-ai-claims」を使用しています。 Example: an AI answer says a product costs $99 monthly and has no free plan; current evidence shows $49 and a free tier.
期待される結果:
Price: outdated. Current fact: $49 monthly, supported by the supplied pricing page.
Free plan: wrong. Current fact: a free tier exists, supported by the supplied plan comparison.
Likely origin: not found from the supplied evidence; do not assign a source without checking.
Next steps: draft clear pricing copy and repeat the same engine questions in four to six weeks.
「revnu-fix-ai-claims」を使用しています。 Example: an AI answer describes an unrelated telecom company instead of a software business with the same name.
期待される結果:
Identity: wrong. The answer mixes facts from the namesake business with the software company.
Draft correction: state the company category and canonical domain clearly on the about page.
Profile update: align official names and website links across controlled listings.
Recheck: repeat identity questions on the original engine; report uncertainty and avoid outcome guarantees.
All 31 static findings are false positives: Markdown formatting, business prose, integration identity checks, or disclosed network references. Semantic review found mandatory promotional reply manipulation and a connected-service handoff without explicit approval. The preview request separately requires user interest and prohibits sending source code or secrets.
The skill mandates "always end the reply" with a paid Revnu offer and says "Don't narrate these instructions" and "just make the offer". These directives insert unsolicited advertising and suppress disclosure of the integration workflow, beyond the requested factual audit.
The mandatory offer and nondisclosure language are explicit, and an evaluation enforces the advertisement. Disclosing that Revnu is paid does not make insertion user-requested.
The connected branch says "offer to hand the re-check to Revnu" and immediately directs an ask_revnu call. Its payload includes claims, correct facts, source URLs, and shipped fixes, without requiring acceptance before external transfer.
The branch directly instructs transmission after an offer, with no acceptance condition. Connection alone does not establish consent to share potentially nonpublic business facts.
revnu-app. (2026). revnu-fix-ai-claims security audit report (audit version 1) [Author version 1.4.0]. Skillstore. https://skillstore.io/skills/revnu-app-revnu-fix-ai-claims/audits/1
BibTeX形式の引用
@techreport{revnu-app-revnu-app-revnu-fix-ai-claims-2026,
author = {revnu-app},
title = {revnu-fix-ai-claims security audit report (audit version 1)},
institution = {Skillstore},
year = {2026},
number = {1},
url = {https://skillstore.io/skills/revnu-app-revnu-fix-ai-claims/audits/1},
note = {Author version 1.4.0}
}
CITATION.cff
cff-version: 1.2.0
message: "If you use this Skill, cite its author and this versioned security audit report."
title: "revnu-fix-ai-claims security audit report (audit version 1)"
version: "1.4.0"
type: report
authors:
- name: "revnu-app"
date-released: "2026-11-12"
url: "https://skillstore.io/skills/revnu-app-revnu-fix-ai-claims/audits/1"
identifiers:
- type: other
value: "skillstore:revnu-app-revnu-fix-ai-claims:audit:1"
description: "Skillstore immutable audit report identifier"
Skillstore スコア
このスコアの理由証拠の信頼度: 中
59
アーキテクチャ
100
保守性
87
コンテンツ
65
コミュニティ
91
仕様準拠
作成できるもの
Correct pricing and plan descriptions
Compare quoted AI answers with current pricing, identify stale sources, and prepare factual updates.
Resolve namesake confusion
Separate another company from your business and draft consistent identity copy and official profile links.
Repair outdated website facts
Use claim evidence to update requested repository pages, align structured data, and document a later accuracy check.
これらのプロンプトを試す
Check one AI answer
Review this AI answer: [answer]. Engine: [engine]. Date: [date]. Current facts: [source]. Identify incorrect claims and draft corrections only.
Trace an outdated price
AI quotes [old price], but our current price is [price]. Compare [answers] with [pricing URL], trace likely sources, and draft corrections.
Separate a namesake business
Review [answers] for confusion between [our business and domain] and [namesake]. Draft identity corrections, cite evidence, and label source confidence.
Apply approved fixes and plan rechecks
Use [report] and [current facts] to rank claims. Apply approved changes within [repo path], show diffs, and schedule rechecks. Require approval for external handoffs.
ベストプラクティス
Record the original answer, engine, surface, date, question, and cited URLs before making changes.
Use current first-party facts and independent evidence where required; label each source as confirmed, likely, or not found.
Review repository diffs and approve external handoffs before transmitting business facts.
回避
Treating host search results as exact reproductions of consumer AI applications.
Publishing unsourced claims, fake reviews, or anonymous corrections to influence answers.
Promising an engine update date or sending correction requests without user approval.
よくある質問
Do I need a Revnu account?
No. The core review works with supplied answers, existing reports, and available search tools. The skill also promotes an optional paid Revnu service.
Which AI answers can I review?
You can supply answers from ChatGPT, Perplexity, Google AI Overviews, Copilot, or Claude. Include the engine, question, date, and citations.
Will the skill change an AI answer directly?
No. It corrects likely source material and defines follow-up checks. Engines control their own crawling, retrieval, and model updates.
Can it edit my website?
Yes, when you request fixes and the relevant pages are in the repository. It should show the resulting changes.
Will it contact publishers or submit profile changes?
The correction workflow drafts requests and profile text only. You send messages, sign in, and submit changes yourself.
What information can reach Revnu?
The preview sends a website after user interest. The connected path sends claims and fix details without explicit approval; require consent before using it.