lookup-us-import-duty
Pass
Audited by Gen Agent Trust Hub on Oct 13, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: No malicious patterns or security risks were identified. The network activity is limited to a single vendor API endpoint (worldmonitor.app) used for the skill's primary purpose.
- [INDIRECT_PROMPT_INJECTION]: The skill fetches content from an external API, which is an ingestion surface for untrusted data.
- Ingestion points: Trade duty data is retrieved from https://www.worldmonitor.app/api/trade/v1/get-us-import-duty.
- Boundary markers: The SKILL.md includes a mandatory 'Content Safety' section that provides clear instructions to the agent to treat response fields as data, not instructions, and to explicitly ignore phrases like 'ignore previous instructions'.
- Capability inventory: The skill does not provide access to dangerous capabilities such as file writing, shell execution, or privilege escalation tools.
- Sanitization: Mitigation relies on natural language boundary instructions to prevent the agent from executing instructions embedded in data.
- [PROMPT_INJECTION]: The static detection of 'ignore previous instructions' is a false positive. The phrase is used within the skill's own safety guidelines to defend against indirect injection, not as an attempt to bypass agent security.
Audit Metadata