Auto-select effort level and model based on task complexity

Problem

Opus 4.6 consumes significantly more tokens than Opus 4.5 due to extended thinking, making it easy to hit 5-hour and weekly usage limits. Users currently have to manually manage token consumption by:

• Toggling thinking on/off (Alt+T)
• Adjusting effort level (low/medium/high)
• Switching models (/model sonnet vs /model opus)

This requires constant manual intervention and context about which setting is appropriate for each task.

Proposed Solution

Let Claude automatically select the appropriate effort level, thinking mode, or model based on the complexity of the incoming task. For example:

• Simple tasks (typo fixes, single-line edits, file reads): low effort or Sonnet
• Moderate tasks (adding a function, small refactors): medium effort
• Complex tasks (architecture changes, multi-file features, debugging): high effort with full thinking

This could work as a new effort level like "effortLevel": "auto" that analyzes the prompt before selecting the appropriate configuration.

Alternatives Considered

• opusplan: Partially addresses this by using Opus for planning and Sonnet for execution, but the split is structural (plan vs execute), not complexity-based.
• Manual switching: Works but adds friction and requires the user to constantly evaluate task complexity themselves.
• Setting a lower default effort: Reduces cost across the board but sacrifices quality on complex tasks that benefit from deeper reasoning.

Additional Context

With usage-based limits (5-hour and weekly caps), efficient token allocation directly impacts how much work users can get done per billing period. An auto mode would let users maximize both quality and throughput without manual tuning.