marketing-pipeline-automation

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

• Third-party content exposure detected (high risk: 0.85). The runtime “Research & Data Crawling” step (e.g., researchTopic with sources like TechCrunch/a16z/Twitter/LinkedIn) fetches and extracts public web/social content authored by outsiders, which is then embedded into the LLM prompt via buildPrompt() → JSON.stringify(config.researchData) → client.messages.create(...).

MEDIUM W013: Attempt to modify system services in skill instructions.

• Attempt to modify system services in skill instructions detected (high risk: 1.00). The skill includes explicit privileged system commands (e.g., "sudo apt-get install chromium-browser" and global installs like "npm install -g pnpm") that instruct obtaining elevated privileges and modifying the host system state.