The Agent Skills Directory

[SAFE]: The skill provides standard, well-documented templates for Node.js, Python, Go, and Rust Dockerfiles using multi-stage builds.
[SAFE]: Promotes security best practices by instructing users to implement non-privileged service accounts (useradd, USER app) and utilizing minimal runtime bases like scratch and Google's distroless images to minimize the runtime attack surface.
[COMMAND_EXECUTION]: Uses bash utilities (docker build, docker images, grep) for the purpose of building containers and validating image size reductions. These are standard operations for a containerization skill.
[INDIRECT_PROMPT_INJECTION]: The skill operates by reading existing project files and Dockerfiles provided by the user to generate optimized versions. Although this introduces a surface for indirect prompt injection from untrusted source code, the risk is inherent to the skill's utility and mitigated by the use of static, standard Dockerfile patterns.
Ingestion points: Reads project files and existing Dockerfiles to determine build steps.
Boundary markers: None explicitly defined for the input interpolation.
Capability inventory: Uses Bash, Write, and Edit to create and modify Dockerfiles.
Sanitization: Relies on the agent's internal logic for template population.

create-multistage-dockerfile