Standards
Pass
Audited by Gen Agent Trust Hub on Nov 1, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill retrieves its content and operational definitions from an external URL (
https://ethskills.com/standards/SKILL.md). It also provides setup instructions to install plugins from a GitHub repository (github.com/austintgriffith/ethskills) and the ClawHub registry. - [INDIRECT_PROMPT_INJECTION]: The skill relies on fetching remote markdown content to guide agent behavior, which presents an attack surface where an attacker could influence the agent by modifying the external source.
- Ingestion points:
https://ethskills.com/standards/SKILL.md(referenced inSKILL.mdandcatalog.json) - Boundary markers: Absent; the skill does not implement delimiters or safety instructions to separate the remote content from system prompts.
- Capability inventory: The skill's demo code utilizes a blockchain SDK (
agent0-sdk) capable of on-chain registration and handling private keys. - Sanitization: Absent; there is no evidence of validation or filtering for the data fetched from the external URL.
- [COMMAND_EXECUTION]: The
catalog.jsonfile specifies an installation command usingcurlto fetch remote content. While the command is used for documentation retrieval, the execution of shell commands during skill installation and the instructions for third-party plugin installation represent significant interaction with external infrastructure.
Audit Metadata