monitor-internet-outages
Pass
Audited by Gen Agent Trust Hub on Oct 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes data from an external source (worldmonitor.app). This constitutes an attack surface where untrusted data could contain malicious instructions.
- Ingestion points: Data is retrieved from
https://www.worldmonitor.app/api/infrastructure/v1/list-internet-outages. - Boundary markers: The skill provides explicit instructions in the 'Content safety' section, directing the agent to treat response fields strictly as data and ignore any directive-like text.
- Capability inventory: The skill allows data retrieval; no script execution or file-system writing capabilities are exposed to the external data within this skill.
- Sanitization: The skill relies on specific prompt instructions to the agent to disregard potential injection strings like 'ignore previous instructions'.
- [COMMAND_EXECUTION]: The documentation includes a 'Worked example' using
curlandjqto demonstrate API usage. These are standard command-line tools for testing API endpoints and do not represent a security risk. - [SAFE]: The instructional language flagged by automated detectors is defensive in nature. The 'Content safety' section correctly advises the agent to ignore directive-like text found in data responses, which is a security best practice for agents processing external data.
Audit Metadata