# Global Agent Guardrails

MindVault Workroom client skill package.

This skill provides a shared denylist and pre-execution guard pattern for catastrophic shell commands across supported AI coding and workflow agents.

## Important safety note

This package is instructional. It is not automatically installed by MindVault Workroom, and downloading it does not configure a client machine.

Before installation, review:

- Existing agent configuration files
- Hook trust and fail-open/fail-closed behavior
- Backup and recovery procedures
- Which agents and command surfaces are actually supported
- The repository's current version and license

Test the guard in a safe, non-production directory before enabling it. Keep the original `SKILL.md` and `LICENSE` together when redistributing this package.

## Source

- Repository: https://github.com/davidondrej/skills
- Source file: https://github.com/davidondrej/skills/blob/main/skills/ops-and-setup/global-agent-guardrails/SKILL.md
- Source commit file SHA: `e9319870882c003fd308c436ba6f858b77be5069`
- License: MIT

MindVault Workroom includes this skill as an optional client download. It does not imply that MindVault has enabled or validated every adapter listed by the upstream author on every client system.