ai-education-data-explorer

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

• Third-party content exposure detected (high risk: 0.72). Required runtime path allows the operating user to supply materialUrl/materialText/materialFile, and the skill’s context composer can ingest that free-form text (including arbitrary outsider-authored content from fetched public pages via materialUrl) into the LLM prompt.

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

• Potentially malicious external URL detected (high risk: 0.80). The provided runner makes runtime POST requests to the remote skill execution API (https://ai-skills.ai/api/execute), so external code/logic on ai-skills.ai is executed at runtime and directly controls the skill's prompts/responses.