write-aida-article-review

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

• Third-party content exposure detected (high risk: 0.75). The workflow can ingest outsider-authored free text when the user supplies an article via a runtime URL or file path, and the agent reads that fetched/downloaded content into the LLM context for “Apply the review.”

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

• Potentially malicious external URL detected (high risk: 1.00). The skill explicitly runs a runtime fetch using curl to load the prompt from https://jeffbailey.us/prompts/aida-article-review/raw.html and then uses that fetched content as the agent's instruction, so this external URL directly controls the agent's prompts.