article-illustrate

CRITICAL E004: Prompt injection detected in skill instructions.

• Potential prompt injection detected (high risk: 0.90). The CONFIRMATION gate (lines 81-83) explicitly tells the agent that "everything after this runs without interruption," which is a directive to ignore or override subsequent user input or interrupts and thus functions like a prompt-injection attempting to change the agent's interactive behavior beyond the skill's stated purpose.

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

• Third-party content exposure detected (high risk: 0.70). The required workflow only instructs the agent to read the “full article” (provided in the prompt or via a user-supplied URL) and then generate illustrations from that text; if the article comes from a runtime-fetched URL or other outsider-authored content, that free-form article prose is ingested into the LLM context, creating an indirect prompt-injection risk.