generate-ors-env

Pass

Audited by Gen Agent Trust Hub on Sep 1, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: No malicious patterns or security vulnerabilities were identified. The skill correctly instructs the agent on how to scaffold and deploy RL environments using the openreward library.
  • [INDIRECT_PROMPT_INJECTION]: The skill scaffolds systems that process external task specifications and tool inputs. It recommends using Pydantic for input validation, which is a standard security practice to mitigate injection risks. * Ingestion points: task_spec and secrets dictionaries in server.py; tool arguments in @tool methods. * Boundary markers: Prompt templates in get_prompt() lack explicit delimiters for untrusted data, though the use of structured blocks (TextBlock, ImageBlock) provides some separation. * Capability inventory: Generated environments can perform network requests, file access, and sandbox management (via E2B integration). * Sanitization: Uses Pydantic BaseModel for tool input schema validation.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 1, 2026, 06:34 PM
Security Audit — agent-trust-hub — generate-ors-env