[Feature Request] Improve context retention and reduce token consumption in agentic workflows

**Bug Description**
first off the model is amazing on what it notices and what it suggests. but it never fails to be missing something obvious or critical. and token usage is way too fast, especially if you overcompensate for it missing things by assigning fable as advisor, then the tokens are burned even faster. most of my work with claude is configuration and spending so much time on alignment consistency and following the rules and actually using the damn skills. just astonishing waste of time and tokens.

**Environment Info**
- Platform: darwin
- Terminal: WarpTerminal
- Version: 2.1.270
- Feedback ID: 59a0cf1f-8732-4c9d-be8a-4081af6de5a0

**Errors**
```json
[]
```