The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes scripts/build-push.sh, a bash script that executes docker build, docker push, and git commands to automate container image workflows. It also includes tests/run.sh, which utilizes standard shell utilities such as mkdir, chmod, mktemp, grep, and cat to perform unit testing and binary mocking.
[INDIRECT_PROMPT_INJECTION]: The skill uses templates to generate orchestration manifests based on user input, creating a potential surface for malicious instructions to be embedded in the output.
Ingestion points: The agent processes assets/Dockerfile.template and assets/docker-compose.template.yml which are designed to be interpolated with untrusted user-provided configuration values.
Boundary markers: Absent. The instructions do not provide explicit delimiters or warnings to the agent to treat user-provided strings as data rather than instructions.
Capability inventory: The skill utilizes Bash and Write tools; the scripts/build-push.sh script performs critical docker operations that could be targeted by maliciously crafted tags or registry names.
Sanitization: Absent. Although the shell scripts use variable quoting to prevent basic command injection, there is no validation or sanitization of the values injected into the generated Dockerfiles or YAML manifests.
[SAFE]: The skill consistently promotes and provides implementations for security best practices, including the use of multi-stage Docker builds to reduce image size, the configuration of non-root users in Dockerfile.template, and the use of Kubernetes NetworkPolicy and ExternalSecret resources in references/k8s-manifests.md.

container-orchestration