skills/openai/plugins/rhea-skill/Gen Agent Trust Hub

rhea-skill

Pass

Audited by Gen Agent Trust Hub on Sep 20, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • General-Purpose Network Tooling: The scripts/rest_request.py utility acts as a generic REST client. It is configured to accept various parameters including URLs, headers, and request bodies. While the skill instructions direct usage toward the Rhea database, the utility's design allows for broader network interactions if provided with different inputs.
  • File System Interaction: The skill includes functionality to store raw response data locally through the save_raw parameter. The script is capable of creating directories and writing text files to a path specified in the input or a default location in the temporary directory.
  • Indirect Prompt Injection Surface: As the skill retrieves and summarizes content from an external biochemical database, it possesses an ingestion point for third-party data. This is a common pattern in search skills that requires the agent to handle potentially untrusted external content.
  • Ingestion points: Data is fetched from rhea-db.org through the REST client script.
  • Boundary markers: The skill does not define specific delimiters for external content within its prompt instructions.
  • Capability inventory: The script uses the requests library for network operations and pathlib for file writing.
  • Sanitization: The script performs basic content compacting by truncating long strings and limiting the depth of nested JSON objects.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 20, 2026, 12:21 PM
Security Audit — agent-trust-hub — rhea-skill