Standards

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.85). The skill’s runtime workflow fetches https://ethskills.com/standards/SKILL.md via curl and instructs the agent to read it, so the LLM ingests outsider-authored free text from public web content.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). The skill's install and setup explicitly fetch remote content at runtime—curl -s https://ethskills.com/standards/SKILL.md (and a plugin install from https://github.com/austintgriffith/ethskills)—and instructs agents to "Read" that URL, meaning externally-hosted content would directly control agent prompts and (in the case of the plugin) install/execute remote code.