web-search
Pass
Audited by Gen Agent Trust Hub on Nov 18, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill instructs the user to install the
beltCLI and related toolkits using theskillspackage manager. These resources are hosted by the inference.sh organization, a platform for AI tool integration. - [INDIRECT_PROMPT_INJECTION]: The skill's primary function is to ingest and process data from the public web via search engines and extraction tools. This creates an inherent attack surface where malicious content on a scanned web page could attempt to influence the agent's behavior.
- Ingestion points: Content fetched via
tavily/search-assistant,tavily/extract,exa/search, andexa/extracttools. - Boundary markers: The provided workflow examples interpolate raw data directly into LLM prompts without explicit delimiters or instructions to ignore embedded commands.
- Capability inventory: The skill utilizes the
beltCLI to run search applications and write results to local files. - Sanitization: There is no evidence of content sanitization or validation performed on the extracted web data before it is passed to LLM models.
Audit Metadata