[FEATURE] Extensible symbol triggers for custom autocomplete (like @ but user-defined)

Summary

Allow users to define custom symbol triggers (e.g., $, #, ~) that invoke user-defined autocomplete behavior, similar to how @ currently triggers file path completion.

Why This Matters

The @ mention for files is one of Claude Code's most useful input features — making it extensible would multiply its value. Users could define their own single-character triggers that surface contextual suggestions relevant to their workflow, without needing to remember and type full slash command names.

This would be a natural extension point that aligns with Claude Code's existing skill/plugin architecture.

Example Use Cases

• Sibling project folders: $ could list other projects in the same parent directory for quick cross-project referencing
• Environment/deployment targets: # could suggest staging, production, dev with associated context
• Team members or reviewers: ~ could autocomplete teammate names for PR assignments or mentions
• Database tables or API endpoints: A symbol could surface schema or route names
• Jira/Linear tickets: A symbol could autocomplete recent or assigned issue keys

Proposed Design

Similar to how hooks run user-defined scripts in response to events, each symbol trigger could invoke a script that returns completion candidates (one per line on stdout). Claude Code would then render these in its built-in autocomplete menu — the same UI used for @ file mentions — so the user can filter and select interactively.

For example in .claude/settings.json:

{
  "symbolTriggers": {
    "$": {
      "command": "ls ../",
      "description": "Sibling projects"
    }
  }
}

The script provides the data, Claude Code provides the picker. This keeps the implementation simple and infinitely flexible — any language, any data source.