worker-module-architecture

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

• Third-party content exposure detected (high risk: 0.90). The worker downloads and executes plugins from the public "Store"/研发商店 via AtomArchiveResourceApi (MarketAtomTask) and pulls code from external repositories (GithubPullTask, CodeGitPullTask, etc.), then executes scripts and reads plugin input/output JSON—clearly ingesting and interpreting untrusted, user-provided third-party content.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill runtime fetches and executes plugin packages via the AtomArchiveResourceApi from the configured gateway/fileGateway endpoint (i.e. the fileGateway/gateway HTTP URL used to download cache/plugin archives), so remote content fetched from that URL is executed locally and therefore poses risk.