This project executes instructions from external config files.
That power comes with responsibility.

• Command steps run arbitrary processes. Treat them as code execution.
• LLM steps may send sensitive data to external providers.
• Configs may be malicious if obtained from untrusted sources.

• Only run configs you wrote or trust.
• Avoid running command steps from unverified sources.
• Keep providers local (e.g., Ollama) for sensitive data.
• If using remote APIs, review their privacy and data-retention policies.
• Run in isolated environments (Docker, venv) when possible.

• Keep jsonschema, pyyaml, and other dependencies updated.
• Use pip-audit or similar tools to scan for vulnerabilities.

If you discover a security issue:

1. Do not open a public GitHub issue.
2. Instead, email the maintainers at [admin@macomber.tech] or use the repository’s Security tab (if enabled).

Future work may include:

• Sandboxing for command steps (e.g. someday it may be cool to consume a docker container as a step).
• Static analyzers for risky templates.