qweather
Pass
Audited by Gen Agent Trust Hub on Mar 10, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill ingests weather data from the QWeather API, presenting a surface for indirect prompt injection where malicious content from the API could influence agent behavior.
- Ingestion points: Weather and location data are fetched via HTTP requests in
scripts/lib/http.tsandscripts/lib/commands.ts. - Boundary markers: No delimiters or instructions are used to distinguish API content from system instructions when the data is presented to the agent.
- Capability inventory: The skill performs network reads from external API hosts.
- Sanitization: The skill handles structured JSON but does not sanitize string fields (e.g., weather descriptions) for embedded instructions.
Audit Metadata