agent-tool-builder

Pass

Audited by Gen Agent Trust Hub on Sep 5, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill serves as a comprehensive educational resource for developing agent tools. It focuses on schema design, error handling, and the use of official SDKs from reputable providers.
  • [INDIRECT_PROMPT_INJECTION]: The skill discusses tools that ingest untrusted data from external sources such as weather APIs and stock tickers. While this represents a potential attack surface, the skill explicitly provides mitigation strategies.
  • Ingestion points: External data sources like weather_api.fetch and get_stock_price (SKILL.md).
  • Boundary markers: Explicitly recommends validation gates and error handling to prevent agents from being poisoned by silent failures.
  • Capability inventory: Uses standard SDK functions for tool calling and result processing.
  • Sanitization: Strictly enforces SQL parameterization and input validation before execution.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 5, 2026, 07:21 AM
Security Audit — agent-trust-hub — agent-tool-builder