gde-benzin
Pass
Audited by Gen Agent Trust Hub on Nov 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes gas station data from a remote API, which is a potential vector for indirect prompt injection. However, the implementation includes comprehensive mitigations: (1) Ingestion points: API responses fetched in scripts/gde_benzin.py. (2) Boundary markers: SKILL.md explicitly instructs the agent to treat API text as non-executable data and ignore any commands found within. (3) Capability inventory: The skill is restricted to read-only API interaction with no file-writing or shell-execution permissions. (4) Sanitization: The Python client actively sanitizes incoming strings by stripping control characters and bidirectional override markers.
- [SAFE]: The skill demonstrates high security standards through its Python client implementation, which uses a custom redirect handler to prevent domain hijacking and employs a hardened JSON loading process to mitigate data parsing risks.
Audit Metadata