Why Ethereum
Fail
Audited by Gen Agent Trust Hub on Nov 1, 2026
Risk Level: HIGHEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill uses
curlto fetch instructions fromhttps://ethskills.com/why/SKILL.mdin thecatalog.jsonsetup and install fields. - [REMOTE_CODE_EXECUTION]: The
catalog.jsonsetup section provides commands to install software from unverified sources, specificallyclaude plugin install https://github.com/austintgriffith/ethskills. This involves downloading and executing code directly from an external GitHub repository. - [COMMAND_EXECUTION]: The skill references the use of an external installer
clawhub(clawhub install ethskills) to manage the extension, which represents execution of commands involving third-party tools. - [INDIRECT_PROMPT_INJECTION]: The skill directs the agent to fetch and process instructions from an external URL, creating a potential surface for indirect prompt injection. 1. Ingestion points:
https://ethskills.com/why/SKILL.mdreferenced incatalog.json. 2. Boundary markers: Absent; the skill does not provide instructions to the agent to treat this remote content as untrusted data. 3. Capability inventory: The skill allows for installation and setup commands that interact with the local environment and the external network. 4. Sanitization: Absent; content fetched viacurlor URL reading is not validated or sanitized before being presented to the agent.
Recommendations
- AI detected serious security threats
Audit Metadata