novel-architect

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

• Third-party content exposure detected (medium risk: 0.30). In SKILL.md the agent ingests outsider-authored free text directly from the user via the interactive interview (e.g., Step 1 core idea), and then includes that text in its drafting/review prompts by reading the created project files (e.g., Synopsis.md/Outline/Characters) without requiring the user to pre-select any trusted document.