lessons-learned
Pass
Audited by Gen Agent Trust Hub on Oct 26, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONNO_CODE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instructions require the agent to search, retrieve, and process data from several external sources, including travel APIs (Seats.aero, Duffel, SerpAPI) and various airline websites. This ingestion of untrusted external content creates a vulnerability surface for indirect prompt injection attacks.\n
- Ingestion points: The agent is instructed to fetch and process search results from Seats.aero, real-time fare data from Duffel, flight prices from SerpAPI, and award availability directly from airline websites like airfrance.com and united.com.\n
- Boundary markers: The skill does not provide instructions for the agent to use boundary markers (like XML tags or specific delimiters) to isolate external data from its own instructional context, nor does it include warnings to ignore instructions found within that data.\n
- Capability inventory: The skill focus is restricted to data retrieval, analysis, and presentation of travel options. The instructions do not involve file-system writes, shell command execution, or other high-privilege operations.\n
- Sanitization: No sanitization, validation, or filtering of the external API responses or website content is prescribed before the data is integrated into the agent's decision-making process.
Audit Metadata