Agent-Safe Commit Guardrails

Version updated for https://github.com/ravisingh11/agent-safe-engineering to version v0.2.0.

  • This action is used across all versions by ? repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

Agent-Safe Engineering is an automated tool designed to ensure AI-generated code changes are understandable and maintainable by humans. It uses human-readable standards, configurable guardrails, focused agent skills, and deterministic validators to enforce constraints on automated code modifications. This ensures that the changes are not only safe but also auditable and reversible.

What’s Changed

Highlights

  • Rejects unknown or misspelled policy parameters with a published strict schema.
  • Explains the resolved policy and included or excluded readability inputs.
  • Generates revision-bound verification receipts for staged trees and Git ranges.
  • Adds a safe dependency-free installer and history-based threshold calibration.
  • Adds hardened repository CI and a reusable GitHub Action pinned at v0.2.0.
  • Self-validates the composite action and its receipt in GitHub Actions.

Validation

  • 27 commit-guardrail and calibration tests passed.
  • 3 installer tests passed.
  • Repository policy copies, schemas, skills, version, and YAML validated.
  • Remote Validate workflow passed on release commit 5f5a973.