Docket – Per-commit evidence records for agent-written code

Docket is a tool that creates per-commit evidence records for code written by AI agents, capturing the implementation journey and verification steps that would otherwise be lost. By folding this journey against the final diff, Docket produces per-hunk evidence records that help reviewers identify which code sections lack evidence of verification. The tool stores records in the repository as an orphan ref and is available as a static binary for macOS, Linux, and Windows.