vulnerability-remediation

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

• Third-party content exposure detected (high risk: 1.00). The skill's required workflow explicitly fetches and analyzes public, user-provided artifacts (e.g., curling registry.hub.docker.com for arigaio/atlas tags, docker pull arigaio/atlas images, and using go get on public Go modules) so untrusted third-party content can be read and directly influence remediation actions.

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

• Potentially malicious external URL detected (high risk: 0.70). The skill issues a runtime curl to https://registry.hub.docker.com/v2/repositories/arigaio/atlas/tags?... and instructs docker pull/run of arigaio/atlas@sha256:, which fetches remote container images (executable code) that the remediation procedure depends on and would be pulled and executed during runtime.