The Agent Skills Directory

[COMMAND_EXECUTION]: The skill optionally executes the docker agent CLI against configuration files to perform a runtime dry-run. This behavior is a documented core capability of the tool and is implemented with security-conscious patterns, such as using subprocess.run with explicit path prefixing to prevent argument injection.
[EXTERNAL_DOWNLOADS]: The skill downloads official JSON schemas and documentation from trusted sources including github.com and docker.github.io. The download implementation is robust, enforcing HTTPS-only connections, response size limits (5MB), and performing digest verification against a local cache.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied YAML configuration files, which constitutes a potential attack surface for indirect prompt injection.
Ingestion points: The scripts/validate_agent_yaml.py script reads and parses YAML configuration files provided by the user.
Boundary markers: The tool enforces strict structure through YAML parsing and JSON schema validation against the official Docker Agent schema definitions.
Capability inventory: The skill can execute the docker agent CLI and curl (as a verified fallback for schema downloads).
Sanitization: The skill contains a dedicated security gate that uses regular expressions to detect and block the inclusion of literal secrets (API keys, tokens, private keys) in the configuration files, requiring the use of environment variable interpolation instead.

docker-agent-builder