The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes a justfile to define and execute shell-based commands for Docker, Kubernetes (Kind/Minikube), and Podman. This provides a structured but broad command execution surface for managing infrastructure tasks.
[DYNAMIC_EXECUTION]: In references/ci-cd.md, a Go helper snippet uses os/exec to programmatically spawn docker compose commands. This is used for creating unique, isolated project environments during E2E testing.
[PRIVILEGE_ESCALATION]: The development configuration in docker-compose.dev.yml includes security_opt: ["seccomp:unconfined"] and cap_add: ["SYS_PTRACE"]. While these settings weaken container isolation, they are documented as necessary requirements for using the Delve debugger in Go development environments.
[CREDENTIALS_UNSAFE]: The skill's templates contain hardcoded default credentials (e.g., POSTGRES_PASSWORD: secret and postgres://test:test@postgres-test:5432/testdb). These are scoped to local development and ephemeral test environments and are consistent with standard developer practices for non-production setups.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to parse and act upon local project files such as justfile and docker-compose.yml, creating a surface where malicious instructions in a repository could influence agent behavior.
Ingestion points: Local configuration files including justfile, docker-compose.yml, and Dockerfile.
Boundary markers: Absent; the skill relies on the integrity of the local workspace files.
Capability inventory: Execution of shell commands via just, kubectl, docker, and podman, as well as process spawning in Go code.
Sanitization: Absent; infrastructure commands are executed directly as defined in the project files.

docker-compose-setup