[FEATURE] Auto-adjust reasoning effort based on prompt complexity

Preflight Checklist

• [x] I have searched existing requests and this feature hasn't been requested yet
• [x] This is a single feature request (not multiple features)

Problem Statement

Currently, users must manually switch effort levels (/effort low, /effort high) depending on task complexity. This creates unnecessary friction — nobody wants to remember to toggle effort before every prompt.

Proposed Solution

Automatically route to the appropriate effort level based on prompt content before the API call is made. This could be a lightweight keyword/heuristic check built into Claude Code itself (no extra model call needed).

Alternative Solutions

• Hook-based approach: UserPromptSubmit hooks can analyze the prompt, but there's currently no API to programmatically change the effort setting from a hook. Exposing effort control in the hook response (e.g. { "effort": "high" }) would also solve this.
• - Manual switching: Current state. Works but adds cognitive overhead.

Priority

Medium - Would be very helpful

Feature Category

Configuration and settings

Use Case Example

| Signal in prompt | Auto effort |
|---|---|
| Stack trace, error log, "bug", "crash", "why isn't this working" | high |
| Architecture, design, multi-file refactor | high |
| "find", "where is", "search", typo fix | low |
| Everything else | medium (default) |

Additional Context

This is especially valuable for workflows that alternate between simple tasks (file search, small edits) and complex analysis (deep bug tracing, architecture decisions). The effort difference in output quality is significant for complex tasks, but wasted on simple ones.