debugging-docker

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

• Third-party content exposure detected (high risk: 0.80). The skill explicitly supports building and inspecting code from arbitrary public URLs (e.g., "docker build ... https://github.com/user/repo.git#branch" and "docker build ... https://example.com/context.tar.gz") and reads Dockerfiles/logs, so the agent will fetch and interpret untrusted, user-generated third‑party content as part of its workflow.