docker-practices

Pass

Audited by Gen Agent Trust Hub on Nov 15, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill references and downloads official Docker base images (e.g., node, postgres, debian, alpine) and established tools like hadolint and Google's distroless images from trusted registries (Docker Hub, Google Container Registry). These references are standard for Docker development and are documented with neutral, informative language.
  • [INDIRECT_PROMPT_INJECTION]: The skill defines an automated validation gate that processes user-provided Docker artifacts. While this creates a potential attack surface if a user provides malicious configuration files, the risk is inherent to the developer tool's purpose and is mitigated by the skill's prescriptive best practices.
  • Ingestion points: The skill reads local Dockerfile, .dockerignore, and compose.yaml files.
  • Boundary markers: Not explicitly defined, but the skill relies on the structured parsing of the underlying tools (hadolint, docker).
  • Capability inventory: The validation process involves running docker build, docker inspect, docker compose, and curl against local containers.
  • Sanitization: The skill employs isolated project naming (PROJ="$(basename "$PWD")-qa") and trap commands to ensure all validation containers and volumes are cleaned up immediately, preventing pollution of the user's environment.
  • [COMMAND_EXECUTION]: The skill includes instructions for running containerized linting and build commands. These commands are executed using standard developer tooling and are targeted at local resources, adhering to the principle of least privilege by advocating for non-root users in runtime environments.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 15, 2026, 04:14 PM