The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the subprocess module to perform administrative tasks required for its operation.
It invokes the Python interpreter to run pip install for its dependencies, ensuring they are localized to the skill's directory.
It spawns ssh processes to create secure tunnels for remote Docker Engine communication, using docker system dial-stdio on the remote host.
All subprocess calls utilize argument lists (not shell strings), which prevents shell injection vulnerabilities.
[PERSISTENCE]: The skill implements a configuration persistence mechanism.
It stores connection parameters, including host addresses and TLS certificate paths, in a local file at .engine/engine.json.
This allows the agent to maintain a consistent connection to a specific Docker environment across multiple interactions.
[DYNAMIC_EXECUTION]: The skill dynamically manages its execution environment to ensure dependency isolation.
It modifies sys.path at runtime to prioritize its local site-packages directory.
It uses importlib.invalidate_caches() and sys.path_importer_cache.pop() to ensure that newly installed packages are correctly recognized by the Python interpreter without requiring a full restart.
[INDIRECT_PROMPT_INJECTION]: The skill acts as a bridge to external Docker environments, creating a potential surface for indirect prompt injection.
Ingestion points: Data retrieved from the Docker Engine API (e.g., container logs, image labels, or status messages) is ingested into the agent's context.
Boundary markers: The skill does not currently implement specific delimiters or 'ignore' instructions to isolate Docker API outputs from the agent's primary instructions.
Capability inventory: Through the Docker SDK, the agent can perform powerful operations such as starting/stopping containers, modifying networks, and reading system logs.
Sanitization: The skill passes raw data from the Docker API to the agent without filtering or escaping potentially malicious instructions embedded in container metadata.

bayma-platform-docker