Function schema required argument grounding

Pass

Audited by Gen Agent Trust Hub on Sep 19, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [SAFE]: The skill is primarily focused on benchmarking and validating tool-call patterns. It does not perform any network operations, file system modifications, or command execution.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user input within its validation logic in validator.py. However, it implements security-conscious heuristics, such as grounding checks, to ensure that the AI agent only uses values explicitly present in the provided text.
  • Ingestion points: The sample object processed by the validate_sample function in validator.py.
  • Boundary markers: The skill's logic enforces grounding, which requires the agent to adhere to the provided dialogue context.
  • Capability inventory: The skill is restricted to string manipulation and regular expression matching. No subprocess calls, network requests, or file writes are identified across any of the scripts.
  • Sanitization: The validator uses tokenization and case-insensitive comparison to verify data integrity without executing the content.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 19, 2026, 05:13 PM
Security Audit — agent-trust-hub — Function schema required argument grounding