web-search
Pass
Audited by Gen Agent Trust Hub on Apr 28, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill references installation scripts and additional components from the vendor's official GitHub repository and website (inference.sh).
- [COMMAND_EXECUTION]: Provides instructions for executing web search and extraction tasks using the 'belt' command-line interface, which is restricted in the YAML configuration to the specific vendor tool.
- [PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection by processing untrusted web content. It follows safety best practices by using boundary markers to delimit external data from instructions.
- Ingestion points: Web search results and extracted page content processed by the 'belt' CLI in SKILL.md.
- Boundary markers: Present; examples use tags such as and to isolate untrusted data.
- Capability inventory: Command execution via the 'belt' CLI tool.
- Sanitization: No explicit filtering or sanitization of web content is described in the prompt templates.
- [SAFE]: No malicious patterns such as obfuscation, credential harvesting, or unauthorized privilege escalation were identified. All external references are to the vendor's own infrastructure.
Audit Metadata