Context window blocking limit too aggressive when auto-compact is disabled

In v2.1.7, the change "Fixed context window blocking limit being calculated using the full context window instead of the effective context window (which reserves space for max output tokens)" causes context to be exhausted at ~80% even when the status bar shows ~20% remaining.

Steps to reproduce

1. Disable auto-compact in settings
2. Use Claude Code until context approaches limit
3. Context blocks at ~80% despite status bar showing remaining capacity

Expected behavior

When auto-compact is disabled, users should be able to use the full context window (as in v2.1.6). The user has explicitly opted out of automatic context management and should have manual control.

Actual behavior

Context blocks at ~80% regardless of auto-compact setting, reserving ~20% for output tokens that may never be needed.

Suggested fix

• Add a setting to disable output token reservation, OR
• Only apply the reservation when auto-compact is enabled (users who disable it expect manual control)

Workaround

Downgrade to v2.1.6: npm install -g @anthropic-ai/claude-code@2.1.6

Environment

• Claude Code version: 2.1.7
• OS: Linux (WSL2)