The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The references/PATTERNS.md file contains a Docker Compose template with hardcoded placeholder credentials, specifically POSTGRES_PASSWORD=password and a DATABASE_URL string. Although intended for local development examples, these values could be inadvertently committed to production systems by an agent following the templates verbatim.
[INDIRECT_PROMPT_INJECTION]: The skill facilitates the creation of Dockerfiles and Compose configurations based on project metadata. This presents a surface for indirect prompt injection where malicious content in processed files (e.g., a poisoned package.json or requirements.txt) could be injected into the generated Docker instructions without sanitization.
Ingestion points: Reads project files (e.g., package.json, requirements.txt) to determine build steps in PATTERNS.md and SKILL.md examples.
Boundary markers: Not present; templates rely on direct interpolation of values.
Capability inventory: The skill uses Write for file creation and Bash (restricted to docker and docker-compose) for execution.
Sanitization: No sanitization or escaping of external content is specified in the skill's instructions.

docker-expert