pets-and-animals-lead-by-18gc2qf
Pass
Audited by Gen Agent Trust Hub on Oct 24, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests data from a local reference file (references/cases.md) and instructs the agent to use tools (browser, image, video) to inspect external content from goodcase.ai and x.com. \n- Ingestion points: The agent reads references/cases.md which contains summaries and links to external case studies. \n- Boundary markers: The workflow includes a "reference contract" (Preserve/Replace/Avoid) and a "Required evidence gate" to structure how external data is used. \n- Capability inventory: The agent is instructed to use browsing and media inspection tools to verify visual fidelity and generates complex video prompts based on that analysis. \n- Sanitization: The skill does not define specific sanitization or validation routines for the data retrieved from external URLs. \n- [DATA_EXPOSURE_AND_EXFILTRATION]: The skill instructions direct the agent to perform network operations by visiting non-whitelisted domains (goodcase.ai, x.com) using available browsing and media inspection tools. \n- Evidence: The workflow step "Inspect its finished media with an available image, video, or browser tool" refers to external URLs listed in references/cases.md.
Audit Metadata