adhx

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

• Third-party content exposure detected (high risk: 0.90). SKILL.md's Workflow explicitly instructs the agent to parse user-provided X/Twitter links and fetch their content from https://adhx.com/api/share/tweet/{username}/{statusId}, causing the agent to ingest public, user-generated social-media posts (tweet/article text) that it then reads and acts on (summarize/analyze), which can enable indirect prompt injection.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill fetches arbitrary post content at runtime from https://adhx.com/api/share/tweet/{username}/{statusId}, and that returned JSON (including article.content) is injected into the model context to drive summarization/analysis, which means external content can directly control prompts.