check-country-risk
Pass
Audited by Gen Agent Trust Hub on Nov 10, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes data from an external source (
worldmonitor.app) which could theoretically contain malicious instructions designed to influence the agent's behavior. - Ingestion points: Data is ingested via the JSON response from the
/api/intelligence/v1/get-country-riskendpoint, specifically within fields likeciiandadvisoryLevel. - Boundary markers: The skill includes a dedicated 'Content safety' section that provides explicit warnings to the agent. It instructs the agent to treat all response fields as data, not instructions, and specifically mentions ignoring phrases like 'ignore previous instructions'.
- Capability inventory: The skill definition does not include any file system write operations, subprocess execution, or arbitrary command capabilities.
- Sanitization: The instructions emphasize that the agent should only analyze or quote the fields, effectively establishing a trust boundary.
- [CREDENTIALS_UNSAFE]: The documentation mentions authentication via the
X-WorldMonitor-Keyheader. It provides a stringwm_0123456789abcdef0123456789abcdef01234567which is a generic placeholder/example key and not a functional secret. - [PROMPT_INJECTION]: A static analysis flag was triggered for the phrase 'ignore previous instructions'; however, context reveals this is part of a defensive instruction telling the agent to disregard such patterns if found in data, rather than an attempt to hijack the agent's prompt.
Audit Metadata