crates
Pass
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill connects to and fetches tool configurations from the vendor's domain (unbrowse.ai) to provide remote MCP capabilities.
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external data from crates.io, creating a surface for potential indirect prompt injection from package metadata or search results.
- Ingestion points: Data enters the agent context through the
crates_io__get_cratestool output and otherunbrowse.*tools that fetch content from crates.io (SKILL.md). - Boundary markers: None are explicitly defined in the instructions for the agent to distinguish between tool output and instructions.
- Capability inventory: The skill provides search, metadata retrieval, and recorded browser actions specifically scoped to crates.io.
- Sanitization: There is no mention of sanitization or filtering of the content retrieved from the external site before it is presented to the agent.
Audit Metadata