press-release-writing

Pass

Audited by Gen Agent Trust Hub on Sep 14, 2026

Risk Level: SAFE
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill references the installation of the belt CLI tool and additional skill modules from the inference-sh and belt-sh GitHub organizations. These references are neutral and part of the documented setup process for the skill's research capabilities.
  • [COMMAND_EXECUTION]: The skill includes instructions for running specific CLI commands using the belt tool (e.g., belt login, belt app run). These commands are used for legitimate research tasks such as market analysis and fact-checking via search assistants.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes external data by querying search tools like Tavily and Exa. This creates an ingestion surface for untrusted data from the web. However, this is a standard feature for search-augmented tools, and there are no signs of unsafe interpolation or missing boundary markers that would escalate this beyond a baseline risk.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 14, 2026, 05:53 PM
Security Audit — agent-trust-hub — press-release-writing