together-hello-world
Pass
Audited by Gen Agent Trust Hub on Sep 10, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill design involves fetching and processing external data from the Together AI model catalog and documentation to ensure models are current.
- Ingestion points: The
WebFetchtool is used to retrieve dynamic content fromdocs.together.aiandapi.together.aito resolve model identifiers. - Boundary markers: The instructions provide explicit guidance to use "sanitized user messages" and set an "explicit output bound" to prevent unintended model behavior.
- Capability inventory: The skill uses
WebFetchfor network communication with the AI provider andWrite/Editfor managing local test files. - Sanitization: The skill contains robust requirements for data handling, specifically instructing the agent to redact secrets in examples (e.g.,
secret=redacted) and forbidding the logging of authorization headers.
Audit Metadata