twist-toolkit

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

• Third-party content exposure detected (high risk: 0.80). This skill fetches user-generated content (threads, comments, direct messages, and search results) from Twist's public API (api.twist.com) as documented in SKILL.md/README and implemented in scripts/twist_api.js, and the AI agent is expected to read and act on that content (summarize, reply, archive), so untrusted third‑party content can materially influence actions despite limited sanitization.

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

• Potentially malicious external URL detected (high risk: 0.70). The skill performs runtime HTTPS requests to Twist (e.g., https://api.twist.com and https://twist.com/oauth/authorize) and returns fetched thread/message content (after limited sanitization) into the AI-facing outputs, meaning external content from https://api.twist.com can directly influence agent prompts and behavior.