together-batch-inference

Pass

Audited by Gen Agent Trust Hub on Mar 4, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill utilizes the official together (Python) and together-ai (Node.js) SDKs to facilitate interactions with the Together AI platform. All network communications are directed to the official API domain api.together.xyz, which is a recognized well-known service.- [PROMPT_INJECTION]: The skill processes user-provided data for batch inference, which presents a surface for indirect prompt injection. Ingestion points: User-defined request payloads within .jsonl files (e.g., batch_input.jsonl) processed in scripts/batch_workflow.py. Boundary markers: No explicit boundary markers or 'ignore embedded instructions' warnings are implemented in the provided scripts. Capability inventory: The skill performs local file system write operations (open, f.write) and network requests to the Together AI API (client.files.upload, client.batches.create). Sanitization: No sanitization, validation, or escaping of the user-provided request content is performed before submission.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 4, 2026, 08:06 PM