fact-checking
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external content (research papers) which creates a surface for indirect prompt injection.
- Ingestion points: The skill uses the
Readtool to ingest content from user-specified file paths in Phase 1 (Claim Extraction). - Boundary markers: The prompt for the extractor agent lacks robust delimiters or specific boundary markers to isolate the paper's content from the agent's instructions, which may allow embedded instructions in a document to override agent behavior.
- Capability inventory: The skill spawns sub-agents with
WebSearchandWebFetchcapabilities and uses theWritetool to create local files based on the processed content. - Sanitization: There is no evidence of sanitization, filtering, or escaping of the claims extracted from the external documents before they are passed to verification agents or written to the final output files.
Audit Metadata