The Agent Skills Directory

[COMMAND_EXECUTION] (HIGH): The skill defines patterns for executing arbitrary commands on the host and inside containers using docker compose exec <service> <command> and docker compose run. This allows for a wide range of host-system influence.
[CREDENTIALS_UNSAFE] (HIGH): Multiple files (docker-compose.yml, Makefile, .env) contain hardcoded default credentials (e.g., MYSQL_ROOT_PASSWORD: root, mysql -u root -proot). While intended for local development, these patterns promote insecure credential management.
[PRIVILEGE_ESCALATION] (HIGH): The 'Troubleshooting' section explicitly includes sudo lsof -i :80 and sudo lsof -i :3306. Directing an agent to use sudo is a high-risk privilege escalation vector.
[INDIRECT_PROMPT_INJECTION] (HIGH): Vulnerability surface identified.
Ingestion points: The skill ingests commands and arguments from the agent's context/user input to pass directly to docker compose exec and the Makefile.
Boundary markers: None identified; commands are interpolated directly into shell strings.
Capability inventory: Host-level shell execution (ls, docker, sudo), container-level execution (bash, drush, composer), and file system modification (chown).
Sanitization: No sanitization or validation of the commands being passed to the containers is present.
[DATA_EXPOSURE] (MEDIUM): The skill provides instructions to access environment variables and project configuration files (.env, settings.php) which may contain sensitive keys or internal network paths.

docker-local