Allow skills to disable or customize @-mention autocomplete in arguments

Resolved 💬 3 comments Opened Apr 14, 2026 by drudolf Closed Apr 14, 2026

Preflight Checklist

What Would You Like?

When typing arguments for a custom skill (e.g. /my-skill @), the @ character triggers the built-in file/folder autocomplete. This is unhelpful when the skill uses @ for its own domain-specific mentions (e.g. @alice, @backend-team).

Requested: A way for skills to control @-autocomplete behavior in their arguments. Either:

  1. A frontmatter option to disable @-autocomplete for a skill (disable-autocomplete: true or autocomplete: false)
  2. A way to define custom completions per skill, so @ suggests skill-defined values instead of files

Use Case

A skill that dispatches work to named roles via @mentions in the arguments:

/my-skill review the auth module @security @backend

Typing @ shows pnpm-lock.yaml, node_modules/, .claude/, etc. — none of which are valid for this skill. The file suggestions clutter the input and cause typos when accidentally accepted.

Proposed Solution

Option 1 (minimal): Add a disable-autocomplete: true frontmatter key that suppresses the built-in @-file autocomplete for that skill's arguments.

Option 2 (richer): Allow skills to define custom completions in frontmatter, e.g.:

completions:
  "@": [alice, bob, backend, frontend, security]

Claude Code Version

Latest

Platform

macOS

View original on GitHub ↗

This issue has 3 comments on GitHub. Read the full discussion on GitHub ↗