deep-research-glim

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

• Third-party content exposure detected (high risk: 0.95). The skill’s required runtime workflow spawns subagents that use glim web search and then full-content extraction (web crawl, GitHub get, Reddit post, Twitter/X, YouTube subtitles), which can ingest outsider-authored free text from public web pages, social posts, and community discussions into the agent’s LLM context.