technical-blog-writing
Pass
Audited by Gen Agent Trust Hub on Apr 14, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses the 'infsh' (inference.sh) CLI tool to perform tasks including authentication, running remote applications, and managing skill extensions.
- [EXTERNAL_DOWNLOADS]: The instructions reference a repository for installing the 'infsh' CLI and adding related skills via 'npx' from the 'inference-sh/skills' GitHub repository.
- [REMOTE_CODE_EXECUTION]: The skill utilizes an external 'python-executor' application to run Python scripts for generating benchmark charts.
- [PROMPT_INJECTION]: The use of a search tool to research blog topics creates an indirect prompt injection surface where untrusted data from the internet could influence the agent's output. Ingestion points: Data retrieved from the 'exa/search' tool as specified in SKILL.md. Boundary markers: None. Capability inventory: The agent can execute shell commands and run Python code via the 'infsh' CLI. Sanitization: No validation or filtering of external search results is described.
Audit Metadata