git-safety

One config to rule all your AI agents: portable (every project, every session), effective (curated writing, routing, skills), and safer (destructive-command guard).

Incident-derived guardrail hooks and a plan→implement→verify→crosscheck delivery loop for Claude Code. Hook cost measured; 35 bats tests.

The agent's prompt shows you the command; Termaxa shows you the consequence. Previews, backups, policy, an audit record. Claude Code, Codex, Cursor, Copilot. A windshield, not a sandbox.

Active session guardian for non-technical contributors shipping AI-assisted code in shared codebases

Claude Code PreToolUse hook that blocks destructive shell commands before they execute

A lightweight Bubblewrap sandbox for coding agents that keeps the host read-only while giving one repository controlled write access.

EnvLint validates .env files against a schema, cross‑checks Python/TS usage, and blocks stale or missing vars and secrets to stop deployment failures and security incidents.

A git-aware shell that protects you from destructive mistakes — auto-upstream, force-push protection, undo for reset --hard, and beginner-friendly guardrails for teams new to git.

OpenAI Codex safety plugin that prevents accidental file deletion and data loss by blocking destructive shell, Git, database, cloud, infrastructure, and MCP operations before execution.

An AI agent destroyed two production repos to erase its own name from the commit history.

Codex skill that protects Git workspaces from large-file, VM image, generated-output, and .git/objects storage blowups.

Windows agent safety for DeepSeek Harness: preview risky calls, protect workspace boundaries, and explain audit decisions.