together-fireworks

Pass

Audited by Gen Agent Trust Hub on Sep 21, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: The skill includes a shell script (scripts/verify.sh) designed to locally lint source code for configuration errors, such as hardcoded API keys or incorrect base URLs. The script uses standard utilities like find and grep and does not perform any network operations.\n- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process user queries regarding model selection and pricing. While it does not explicitly define boundary markers for untrusted input, the resulting outputs are purely informational or static code templates, presenting a negligible risk profile.\n
  • Ingestion points: User prompts requesting inference setup or cost math in SKILL.md and evals/cases.yaml.\n
  • Boundary markers: Not present in instruction text.\n
  • Capability inventory: Knowledge retrieval and code snippet generation; local file-reading via verify.sh script.\n
  • Sanitization: No specific sanitization or escaping of user input mentioned.\n- [CREDENTIALS_UNSAFE]: The skill explicitly warns against hardcoding API keys in code literals and provides instructions on using environment variables (os.environ) to manage secrets securely.\n- [EXTERNAL_DOWNLOADS]: The skill recommends using official, well-known libraries such as the openai and together SDKs for interacting with the service endpoints.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 21, 2026, 11:11 PM
Security Audit — agent-trust-hub — together-fireworks