ClaudeHack
0.000

Set disable-model-invocation on skills with side effects

Terminal / Productivity / Cost Optimization·13 hours ago·verified 13 hours ago·easy·~2 min setup·verified

Mark deploy/ship/destructive skills so only an explicit /command runs them—Claude cannot auto-trigger them mid-session.

The problem

Skills with deploy, delete, or release side effects can be auto-invoked when Claude thinks they are relevant.

The hack

Add `disable-model-invocation: true` to the skill frontmatter for any command that mutates production or shared state.

Why it works

User-invocable skills can still be model-triggered unless you opt out. Explicit `/ship` is safer than Claude deciding “now seems like a good time to deploy.”

Setup

1. Open the skill’s `SKILL.md`. 2. Add `disable-model-invocation: true` to frontmatter. 3. Keep `user-invocable` true so `/ship` still appears in the menu. 4. Test by asking Claude to deploy in natural language—it should not auto-run the skill.

Code

---
name: ship
description: Deploy the current branch after checks pass
disable-model-invocation: true
---
Run the release checklist, then deploy only if green.

Additional details

Use this for `/ship`, `/deploy`, `/drop-db`, and similar irreversible workflows.

Source

blog·View original →·by arps18

Does this still work?

0 said it works · 0 said it's broken

Related

Discussion

No comments yet. Be the first to say whether this worked for you.