lofi-monk/lofi-gate

Overview

This skill enforces a mandatory pre-commit checkpoint that verifies changes stayed within scope and that tests were not tampered with. It structures a concise, machine-readable judgment so agents and CI pipelines can automatically accept or reject changes. Use it to prevent test manipulation, scope creep, and accidental overreach before pushing code.

How this skill works

The skill compares staged or committed diffs against the original mission instructions, checks for modifications to existing tests, skipped or weakened assertions, deleted tests, and scope violations. It emits a clear PASS/FAIL judgment with a specific reason and required fix, and provides commands to log the verdict for auditability. Outputs are intentionally strict and token-efficient to avoid agent hallucination.

When to use it

• Before committing or pushing changes to a repository
• As a mandatory CI quality gate step
• When you want to self-audit changes against a defined mission or task
• When preventing test tampering or unauthorized scope expansion
• In automated agent workflows to reduce hallucination risk

Best practices

• Stage changes and save git diff output as evidence before running the check
• Keep the original mission instructions accessible (conversation history, implementation_plan.md, task.md)
• Do not alter existing test files unless the mission explicitly requires it
• Address each failing question precisely and rerun the check after fixes
• Log the verdict using the provided logger command for traceability

Example use cases

• Local pre-commit hook that rejects commits flagged as REJECTED by the checkpoint
• CI pipeline step that blocks merge if the skill returns a FAIL judgment
• Agent workflow that must confirm adherence to a task before proceeding to next step
• Developer running a final compliance pass to ensure tests haven’t been weakened
• Audit trail generation by recording PASS/FAIL with the built-in logger