The Agent Skills Directory

[COMMAND_EXECUTION]: The skill employs local shell commands and the Docker CLI (e.g., docker version, docker ps, find, docker build) to inspect the host environment and validate container configurations. These operations are restricted to the local context and are necessary for the skill's primary function.\n- [PROMPT_INJECTION]: Instructions are strictly technical and persona-aligned, with no evidence of attempts to bypass safety filters or override agent behavior. The skill was evaluated for Indirect Prompt Injection vulnerability surface: Ingestion points include local Dockerfiles and Compose files identified via find; Boundary markers are absent; Capabilities include docker build, run, and exec; Sanitization is absent. Given the intended use-case of a container expert, this is a standard risk profile.\n- [DATA_EXFILTRATION]: No network operations to external or untrusted domains were identified. External communications are limited to well-known health check patterns on localhost or official Docker tools like Scout.\n- [EXTERNAL_DOWNLOADS]: No remote scripts or packages are downloaded and executed. All external references in Dockerfile examples target well-known and official base images from trusted registries (e.g., Docker Hub, Google Container Registry).

docker-expert