sealos-deploy

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 reads arbitrary GitHub repositories (modules/pipeline.md / modules/preflight.md Phase 1 and Step 2/3) and queries public registry pages/APIs and README contents (modules/pipeline.md Phase 2: detect-image.mjs, curl checks against Docker Hub and GHCR), and those untrusted, user-generated sources are parsed and used to decide builds, registry choice, image pull-secret creation, and deployment actions — enabling indirect prompt-injection via third-party content.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill's preflight instructions explicitly run a remote installer at runtime (e.g., "curl -fsSL https://get.docker.com | sh") to fetch-and-execute code when installing Docker, which is remote code execution performed by the agent if the user consents.