The Agent Skills Directory

Demonstration Credentials: The file assets/compose-web-app.yaml contains hardcoded credentials such as POSTGRES_PASSWORD: secret and a DATABASE_URL with embedded secrets. These are explicitly labeled as "Demo only" and are intended for educational purposes. The skill's core guidance and the assets/bad-vs-good.md file provide clear remediation steps, recommending the use of environment files (.env) and Docker secrets for production environments to avoid hardcoding sensitive information.
Local Configuration Validation: The skill includes a shell script, scripts/verify-compose.sh, which runs docker compose config to validate the syntax of generated or modified Compose files. This is a standard, benign development utility used to ensure configuration correctness.
Indirect Prompt Injection Surface: The skill is designed to process user-provided compose.yaml files. It includes the capability to execute the docker compose config command for validation purposes. While this represents a surface for indirect prompt injection if malicious instructions were embedded in the configuration data, the skill includes detailed verification runbooks in checks/verification.md for manual review and assumes standard guardrails are in place during agent interaction.

docker-compose-patterns