openai-ai-integration
Pass
Audited by Gen Agent Trust Hub on Aug 20, 2026
Risk Level: SAFE
Full Analysis
- [CREDENTIALS_UNSAFE]: The skill provides strong guidance on credential safety, explicitly instructing developers to never log API keys and to use environment variables or configuration files for secret management. It correctly identifies the API key as a sensitive boundary that should stay within the SDK client constructor.
- [EXTERNAL_DOWNLOADS]: Recommends the official 'openai' SDK for Python and Node.js. These are well-known, official libraries from a trusted service provider.
- [DATA_EXPOSURE]: Includes explicit warnings against data exposure through logging. It recommends constructing log payloads from typed request objects that intentionally omit credentials.
- [INDIRECT_PROMPT_INJECTION]: The skill acknowledges that LLM outputs are untrusted network boundaries. It mandates schema validation on all responses and provides a resilient parsing strategy to handle cases where models return malformed or prose-wrapped JSON.
- Ingestion points: LLM API responses (completion content).
- Boundary markers: Explicitly recommends validation on the response even when using strict mode.
- Capability inventory: Interactions are limited to SDK chat completion calls.
- Sanitization: Recommends decoding and checking using schema-validation tooling.
Audit Metadata