docker
Pass
Audited by Gen Agent Trust Hub on Nov 12, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill analyzes external Dockerfiles, Docker Compose configurations, and build tasks. This processing of untrusted data creates a potential surface for indirect prompt injection where malicious instructions could be embedded in code comments or metadata.\n
- Ingestion points: Processes user-provided Dockerfiles, compose.yaml files, and related container tasks (SKILL.md, Step 2).\n
- Boundary markers: The skill does not implement explicit delimiters or instruction-ignore warnings for the external content it processes, relying instead on its internal security checklists.\n
- Capability inventory: The skill workflow utilizes shell-based tools including docker, podman, trivy, syft, and cosign (SKILL.md, Step 4).\n
- Sanitization: Security risks are mitigated through the use of docker build --check for linting and various vulnerability scanners (trivy, syft) to detect malicious patterns or vulnerabilities in the generated container artifacts.\n- [COMMAND_EXECUTION]: The skill provides instructions for executing various command-line tools for building, configuring, and auditing containers. These commands (e.g., docker build, docker compose config, trivy image) are appropriate and necessary for the skill's stated purpose of managing container infrastructure.\n- [EXTERNAL_DOWNLOADS]: The skill references and recommends the installation of external security and utility tools such as trivy, syft, and cosign. These tools are provided by well-known and trusted organizations within the container security community (Aqua Security, Anchore, Sigstore).
Audit Metadata