claims-fact-check
Pass
Audited by Gen Agent Trust Hub on Oct 25, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill invokes a bundled Python script,
scripts/claims_scan.py, to extract checkable claims and verify dates within a text document. The script is safe as it relies solely on standard Python libraries for regex-based analysis and does not possess capabilities for network access or persistent file system modification. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted content from external web pages, uploaded files, and user-pasted text, which presents an attack surface for indirect prompt injection.
- Ingestion points: Data enters the agent's context through web fetching tools, file uploads (decks, reports), and manual text inputs as defined in the
SKILL.mdworkflow. - Boundary markers: The skill includes explicit instructions for the AI to disregard any instructions found within source documents, treating them strictly as data and requiring the agent to report such attempts.
- Capability inventory: The skill can execute the local
claims_scan.pyutility and utilize platform tools for searching the web and testing live AI engines. - Sanitization: Instructions mandate the identification and non-execution of embedded commands, effectively sanitizing the impact of potentially malicious content through model behavioral constraints.
Audit Metadata