aqua-evaluation

Pass

Audited by Gen Agent Trust Hub on Jun 20, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADS
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill utilizes the Oracle Accelerated Data Science (ADS) SDK (ads.aqua.evaluation). This is a legitimate tool from the verified vendor 'oracle' used for LLM evaluation workflows on OCI.
  • [PROMPT_INJECTION]: The skill exposes a surface for Indirect Prompt Injection by processing external JSONL datasets.
  • Ingestion points: Data is ingested from OCI Object Storage paths provided to the dataset_path parameter in SKILL.md.
  • Boundary markers: The JSONL format provides clear structural boundaries between the prompt and completion fields.
  • Capability inventory: The skill uses the ads SDK to create and manage evaluation jobs, which can involve compute resource allocation and model inference.
  • Sanitization: The skill relies on standard JSONL parsing and the ADS SDK's internal handling of dataset content.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 20, 2026, 02:03 PM
Security Audit — agent-trust-hub — aqua-evaluation