The Agent Skills Directory

Hardcoded Development Credentials: The assets/compose-dev.yaml template includes a default fallback password (apppass) and hardcoded connection strings (postgres://appuser:${POSTGRES_PASSWORD:-apppass}@db:5432/appdb) to enable one-command startup for local development. While the documentation in SKILL.md labels these as fallbacks and recommends using .env overrides, hardcoded credentials in configuration templates are a security pattern to monitor.
Automated Command Execution: The scripts/verify-setup.sh script executes docker compose config --quiet to validate the syntax of generated configuration files. This is a standard verification tool used to ensure project readiness but involves the automated invocation of the Docker CLI.
Indirect Prompt Injection Surface: The skill processes user-provided project requirements to generate specific configuration files, which presents a surface for potential instruction interpolation.
Ingestion points: User project descriptions and service requirements are used to populate templates in assets/ and directives in SKILL.md.
Boundary markers: The skill provides a comprehensive .dockerignore example to limit context exposure and instructs the use of BuildKit secret mounts for registry credentials.
Capability inventory: The skill facilitates the creation of Dockerfile and compose.yaml files, as well as the execution of Docker build and orchestration commands.
Sanitization: The guidance explicitly warns against copying credential files or using ARG/ENV for secrets, promoting BuildKit secrets as a more secure alternative.

docker-project-foundations