Per-agent Tab completion scoping via rules and agent config

Feature request for product/service

Cursor IDE

Describe the request

Tab completions have a single global scope — on or off. There is no way to automatically
disable them when a Socratic mentoring agent is active, or re-enable them when switching
to normal editing.

Manually toggling editor.inlineSuggest.enabled and cursor.completions.enabled in
settings.json via an agent command is fragile and breaks flow on every context switch.

Allow .cursor/rules/*.mdc or .cursor/agents/*.md to set completion behavior when
that agent/rule is active:

# In a mentoring agent or rule
editor:
  inlineSuggest: false
  cursorCompletions: false

When the rule/agent is active, completions are suppressed. When switching away, they
revert to the user’s settings.json defaults. Or at minimum, mode-based profiles:

{
  "cursor.completions.modes": {
    "agent": true,
    "ask": false,
    "plan": false
  }
}

Use case: University C++ course with a Socratic mentoring agent hierarchy —
completions off during guided learning, on during code hardening and security review.

Prior art:

Happy to test a build.