hosted-agents

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

• Third-party content exposure detected (high risk: 0.90). The skill explicitly clones and loads repository code into sandboxes (e.g., git clone https://x-access-token:{token}@github.com/{repo} in the Modal/ImageBuilder examples), processes Slack thread messages for repository classification, and extracts DOM from arbitrary web pages via the Chrome extension content script—all of which ingest untrusted, user-generated third-party content that the agent is expected to read and act on.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill performs runtime git clones from GitHub (e.g. "https://x-access-token:{token}@github.com/{repo}") and then runs build/install/test/dev commands inside the sandbox, so externally fetched repository content is retrieved at runtime and can execute remote code.