Currently, it seems that if I want to disable skill usage, I need to do it individually in the frontmatter using the disable-model-invocation option for each skill.
Is there any built-in way to toggle skill usage dynamically in bulk—either enabling or disabling all skills at once—rather than manually updating each one separately?
If not currently supported, are there any recommended workflows or configurations to achieve a similar effect more efficiently?
There isn’t currently a built-in way to toggle all skills at once in Cursor. disable-model-invocation: true is per-skill frontmatter. It prevents the agent from auto-applying a skill, but you can still invoke it manually with /skill-name (when it works as documented).
Practical workarounds:
Script - batch-add disable-model-invocation: true to all SKILL.md files under .cursor/skills/, .claude/skills/, and ~/.cursor/skills/.
Rename the folder - e.g. .cursor/skills → .cursor/skills.disabled to stop discovery entirely without editing each file.
Rename files - SKILL.md → SKILL.md.disabled (agents only discover files named exactly SKILL.md).
Use paths - scope skills to specific file globs so they don’t surface on unrelated work.
Note: Cursor’s built-in skills in ~/.cursor/skills-cursor/ aren’t user-editable, so there’s no bulk off switch for those. A global “skills master toggle” would be a reasonable feature request.
Hey, @helldark’s reply above covers it correctly. There’s no built-in bulk toggle right now, skill control is per-skill only via disable-model-invocation in the frontmatter. Practical workaround: rename the .cursor/skills folder to .cursor/skills.disabled, or rename the individual files from SKILL.md to SKILL.md.disabled. That’s the fastest way to disable everything at once without editing the frontmatter in each file.