case-study-writing
Pass
Audited by Gen Agent Trust Hub on May 19, 2026
Risk Level: SAFE
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes a vendor-provided CLI tool (
belt) to perform tasks such as user login and running specialized applications for web search and data processing. These operations are within the expected scope for the skill's functionality. - [EXTERNAL_DOWNLOADS]: The documentation references installation scripts and additional skills hosted on the official GitHub repository for the vendor (
inference-sh/skills). These are recognized as trusted vendor resources. - [REMOTE_CODE_EXECUTION]: The skill demonstrates the use of a Python executor app (
infsh/python-executor) to generate data visualizations (charts) usingmatplotlib. The code provided in the examples is static, task-specific, and utilizes standard libraries for its intended purpose. - [INDIRECT_PROMPT_INJECTION]: The skill creates an attack surface by ingesting data from external search tools (
tavily/search-assistant,exa/search) to inform the case study writing process. - Ingestion points: Search results from external providers (SKILL.md).
- Boundary markers: None explicitly defined for tool outputs.
- Capability inventory: File system writes (chart generation), network access (via research tools), and Python execution (via vendor executor).
- Sanitization: No explicit sanitization of research data is described. This is a common characteristic of research-oriented writing skills.
Audit Metadata