univer-team-standup

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

• Third-party content exposure detected (high risk: 0.80). The Auto workflow's "Evidence sources" explicitly includes "Git and GitHub: commits, branches, PRs, issues, review comments, CI results" (Workflow: Auto / Collection rules), which are public/user-generated third-party contents that the agent must read and summarize and which directly influence priority, status, and write/submit decisions.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill explicitly bootstraps required code at runtime (e.g., npx skills add dream-num/skills which pulls https://github.com/dream-num/skills and npm install -g univer-cli@latest which installs remote npm package code) and those fetched repositories/packages are required dependencies that will install and run code locally, so they can execute remote code used by the agent.