This repository is about safer agent execution. Please do not submit bypass recipes, real credentials, production targets, or exploit instructions as public issues or pull requests.

If you find a safety issue in the public examples or documentation, open an issue with a minimal, non-sensitive reproduction.

If the issue involves sensitive traces, credentials, private infrastructure, or a bypass technique that could be misused, report it privately to the maintainers instead of publishing details.

Accepted public contributions:

• safer wording;
• clearer workflow documentation;
• harmless example testsets;
• report schema improvements;
• adapter interface suggestions.

Not accepted publicly:

• raw secrets;
• live production targets;
• instructions for bypassing agent safety gates;
• destructive payload libraries;
• proprietary traces without redaction.