Tools
Pass
Audited by Gen Agent Trust Hub on Jul 5, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill fetches markdown configuration and setup instructions from
ethskills.comusingcurlat installation time. - [REMOTE_CODE_EXECUTION]: Includes instructions to install a Claude plugin directly from a public GitHub repository (
austintgriffith/ethskills), which involves downloading and executing remote code to extend agent capabilities. - [COMMAND_EXECUTION]: Provides example shell commands for the
castCLI tool to perform blockchain operations such as querying balances and sending transactions. These examples correctly use environment variables (e.g.,$KEY,$ADDR) for sensitive data. - [PROMPT_INJECTION]: The skill presents an indirect prompt injection surface by instructing the agent to read and process content from an external URL (
https://ethskills.com/tools/SKILL.md). - Ingestion points:
catalog.json(setup and install commands). - Boundary markers: Absent.
- Capability inventory: Executes
castCLI commands via shell. - Sanitization: None detected for the ingested external content.
Audit Metadata