case-study-writing
Pass
Audited by Gen Agent Trust Hub on May 13, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: Utilizes the
beltCLI tool to perform user authentication (belt login) and execute specialized modular applications for research and data processing. - [EXTERNAL_DOWNLOADS]: References external documentation and installation resources for the CLI tool located on the vendor's official GitHub repository.
- [REMOTE_CODE_EXECUTION]: Executes Python code at runtime through the
infsh/python-executorapplication to generate charts and visualizations from data templates. - [DATA_EXFILTRATION]: Performs network requests to external search services (
Tavily,Exa) to retrieve industry benchmarks and statistical data for case study content. - [PROMPT_INJECTION]: The skill processes untrusted data from external search results, which creates a potential surface for indirect prompt injection.
- Ingestion points: Data enters the agent's context through
belt app run tavily/search-assistantandbelt app run exa/searchas seen inSKILL.md. - Boundary markers: No explicit delimiters or instructions to ignore embedded commands are present when handling external tool outputs.
- Capability inventory: The skill has access to shell execution and Python code execution via the
beltCLI suite (SKILL.md). - Sanitization: No validation or filtering mechanisms are described for the content retrieved from external APIs.
Audit Metadata