Stacks Documentation
SkillsChangelog
TeamSponsorsPartnersPostcardwareAwesome StacksContributing

Guard

Copy page
Copy page as Markdown for LLMs
View as Markdown
View this page as plain text
Open in ChatGPT
Copy content and open ChatGPT
Open in Claude
Copy content and open Claude
Open in Perplexity
Search with Perplexity AI
On this page 5

stacks-guard · Engineering craft · model-invoked

Safety rails, in two layers. The catalogue tells the agent which commands are catastrophic, which merely warrant a warning, and which are worth noting. The bundled PreToolUse hook makes it enforcement rather than advice, blocking the unrecoverable ones before they execute while leaving rm -rf node_modules and a local migrate:fresh alone.

When to reach for it

  • Detecting destructive commands (rm -rf, DROP TABLE, force-push, git reset --hard, migrate:fresh against production)
  • Installing a PreToolUse hook that blocks them before they run
  • Freeze mode for focused debugging
  • A pre-commit safety scan

Inside the skill

The sections an agent reads once the skill loads.

  • Destructive Command Detection
  • Freeze Mode
  • Pre-Commit Safety
  • Stacks-Specific Guards
  • Make it enforcing
  • Rules

Supporting files

Reference and scripts the skill reaches for on demand, rather than loading up front.

Using it

This one is model-invoked. Your agent reaches for it on its own when the task matches, and you can also call it by name:

/stacks-guard

Source: stacks-guard/SKILL.md. Shadow it for one project with app/Skills/stacks-guard/SKILL.md, then re-run buddy setup:ai. See Writing your own.