Testing
Pass
Audited by Gen Agent Trust Hub on Nov 1, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill uses
curlto fetch instructional content fromhttps://ethskills.com/testing/SKILL.mdas part of its installation and setup process. - [EXTERNAL_DOWNLOADS]: The
catalog.jsonfile references external installation commands for plugins hosted on GitHub (https://github.com/austintgriffith/ethskills) and through theclawhubregistry. - [INDIRECT_PROMPT_INJECTION]: The skill directs the agent to read instructions from a remote, external URL (
https://ethskills.com/testing/SKILL.md). This creates a dependency on an external resource that could be modified to influence agent behavior. - Ingestion points: The
setupinstructions and theSKILL.mdbody both point to the remote URL as a source of information for the agent. - Boundary markers: Not present; the instructions do not wrap the external content in delimiters or provide warnings to ignore potentially malicious embedded instructions.
- Capability inventory: The skill facilitates smart contract testing, which involves running developer tools (Foundry) and interacting with local codebases.
- Sanitization: Not present; the remote markdown content is consumed without validation or filtering.
Audit Metadata