hotel-concierge
Pass
Audited by Gen Agent Trust Hub on Sep 29, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process variable guest information, creating a surface for indirect prompt injection.\n
- Ingestion points: User-supplied arguments describing guest profiles (nationality, age, interests) and complaint scenarios (levels 1-3) in SKILL.md and templates/prompt-main.md.\n
- Boundary markers: The prompt templates in templates/prompt-main.md lack explicit delimiters or markers to isolate user-provided data from the rest of the instructional context.\n
- Capability inventory: The skill is strictly limited to text generation; it has no access to network operations, file system modifications, or shell command execution.\n
- Sanitization: No sanitization or filtering logic is present to handle potentially malicious instructions embedded within guest profiles or complaint descriptions.
Audit Metadata