disabled-skill

To test skill workflows without invoking external models, enable the disable-model-invocation flag in SKILL.md. This runs the skill in a safe mode that prevents live AI calls, allowing you to validate prompts and logic without incurring costs or side effects.

The best way to validate prompts in CI without live AI calls is to use a skill with the disable-model-invocation flag enabled. This ensures risk-free experimentation in staging environments by preventing external model invocations during dry runs.

You can toggle between normal and safe testing modes by modifying the disable-model-invocation flag in the SKILL.md frontmatter. This allows you to switch to a mode that prevents any external model invocation for QA and dry runs.

Disabling model invocation does not prevent prompt and flow validation in development pipelines. It specifically blocks external model calls so you can safely test logic and validate workflows without side effects or unexpected costs.

Yes, you can reuse skills in staging without side effects by enabling the disable-model-invocation flag. This operates the skill in a controlled mode that prevents unintended model invocations during testing and QA.

You should disable model invocation during QA testing to prevent unintended live AI calls. This allows you to validate prompts, logic, and flows in a controlled execution environment without introducing side effects or incurring costs.