ai-reviewed-commit

Automating pre-commit code review for staged git changes involves routing modifications through a model-based review or debate loop. The process enforces AGENTS.md rules, applies a 3-round fix-and-retry cycle, and generates a Conventional Commits message for the final commit.

Yes, AI can generate Conventional Commits messages automatically. After validating staged changes against AGENTS.md rules and resolving issues through a fix-and-retry loop, the orchestrator produces a compliant Conventional Commits message for the final commit.

Yes, AI pre-commit review distinguishes between self-authored changes and changes by others. It routes self-authored modifications to a standard model review, while routing changes by others to a debate for more rigorous evaluation before committing.

A 3-round fix-and-retry loop in an AI code review workflow is a strict cap for resolving identified issues. The orchestrator attempts to fix review findings and retries the review up to three times, ensuring issues are resolved before the commit proceeds.

You can enforce AGENTS.md compliance checks before a git commit by routing staged changes through an automated orchestrator. The process validates modifications against AGENTS.md rules during the review loop, ensuring governance before generating the final commit message.

You should not use an automated pre-commit review loop when changes are not yet staged, as the workflow requires staged inputs to function. Additionally, the orchestrator will not work unless SA-mode propagation is explicitly honored with --sa-mode true on CSA commands.