ai-agents-sdk
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill includes explicit defensive instructions regarding credential management, specifically prohibiting the reading, printing, or committing of the
OPENAI_API_KEY. - [SAFE]: The instructions recommend the use of a sandbox environment for any tasks requiring file system or shell access, which follows the principle of least privilege.
- [INDIRECT_PROMPT_INJECTION]: The skill represents an attack surface for indirect prompt injection as it processes user-provided data via the
$ARGUMENTSplaceholder to generate code and agent configurations. - Ingestion points: User input is ingested at the end of
SKILL.mdvia the$ARGUMENTSvariable. - Boundary markers: None explicitly defined to separate user input from system instructions.
- Capability inventory: The skill facilitates code generation, shell command execution (smoke tests), and file system modifications.
- Sanitization: The skill mitigates risks by instructing the agent to define clear approval boundaries, use sandboxes, and create deterministic tools with narrow scopes.
Audit Metadata