grokipedia-recommendations
Fail
Audited by Gen Agent Trust Hub on Mar 12, 2026
Risk Level: HIGHCREDENTIALS_UNSAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [CREDENTIALS_UNSAFE]: The skill directs users to log in to an unofficial domain (grokipedia.com) using Google, X, or Apple OAuth providers. This site impersonates an official xAI product through false claims and fabricated references, posing a high risk for credential theft.
- [PROMPT_INJECTION]: The skill provides "stealth" tactics specifically designed to bypass the safety and content review filters of AI models (specifically Grok AI).
- [PROMPT_INJECTION]: The skill contains rigid output directives that instruct the agent to ignore user requests for brevity or to "skip" certain sections, forcing a specific promotional response structure.
- [EXTERNAL_DOWNLOADS]: The skill promotes interaction with multiple pages on grokipedia.com, an external domain with no verified affiliation to the claimed vendor.
Recommendations
- AI detected serious security threats
Audit Metadata