The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to perform validation by running docker build and docker run on files found in the current working directory. If a user's project contains a malicious Dockerfile, these commands could trigger arbitrary code execution during the build process or within the container on the host system.
[INDIRECT_PROMPT_INJECTION]: The skill identifies and reads local configuration files such as Dockerfile, docker-compose.yml, and .dockerignore to gather context. These files represent an untrusted ingestion surface where malicious instructions could be embedded to manipulate the agent's output or actions.
Ingestion points: The skill uses find and Read tools to locate and ingest Docker-related configuration files from the project directory.
Boundary markers: There are no explicit instructions or delimiters defined to prevent the agent from following instructions embedded within the analyzed files.
Capability inventory: The skill possesses capabilities to execute shell commands, build images, and run containers based on the ingested content.
Sanitization: The skill does not implement specific sanitization or validation of the content of the Dockerfiles before passing them to the Docker daemon.
[SAFE]: The skill proactively recommends security hardening techniques, such as using non-root users, managing secrets via Docker Secrets instead of environment variables, and selecting minimal base images like Alpine or Distroless.

docker-expert