web-search
Pass
Audited by Gen Agent Trust Hub on Mar 16, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill requires permission to execute the
infshcommand-line interface. This tool is used to perform web searches, extract webpage content, and interact with various AI models on the vendor's platform. - [PROMPT_INJECTION]: There is an indirect prompt injection surface because the skill ingests content from external, untrusted websites and interpolates it into LLM prompts.
- Ingestion points: External web data enters the agent context via the
tavily/extractandexa/searchtools as shown in the workflow examples. - Boundary markers: None are present in the provided workflow examples; extracted content is placed directly into prompts using simple placeholders like
<search-results>and<content>. - Capability inventory: The skill utilizes the
Bash(infsh *)tool, which allows for a wide range of operations including network requests (via the platform) and interaction with multiple LLM providers. - Sanitization: No evidence of content sanitization or filtering (such as stripping HTML comments or instruction-like text) is found in the skill definition.
Audit Metadata