project-launch

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

• Third-party content exposure detected (high risk: 0.90). This skill's Phase 6 workflow ("How Claude Code Researched Outreach Targets" and the checklist items like "Run 4-wave parallel web search") explicitly instructs the agent to fetch and read user-generated public content (GitHub issues/discussions, Reddit, Stack Overflow, blog comments) and to act on those findings (categorize targets, generate outreach and post replies), which clearly exposes the agent to untrusted third-party content that could carry indirect prompt-injection instructions.