What problem does it solve? Teams using Sixth Sense need a consistent, project-level way to control which AI models subagents use and how much reasoning effort they spend, instead of re-deciding on every task dispatch. ## Core Features & Use Cases - Guided Setup Interview: Walks the user through five questions covering authorized models (Sonnet, Haiku), default effort level, Haiku batching, Opus second-opinion review, and Opus routing for the hardest tasks. - Persistent Configuration File: Generates a .sixth-sense/model-preferences.md file with YAML frontmatter that is read automatically by dispatching-parallel-agents on every dispatch. - Cost vs. Capability Control: Lets teams trade off cost (Haiku batching, Sonnet-only mode) against reasoning depth (Opus review passes, opus-high escalation tier). - Use Case: A team lead starting a new project runs the setup once, answers the five questions with recommended defaults, and every subsequent parallel agent dispatch in that project respects the saved model and effort preferences. ## Quick Start Run the configuring-sixth-sense skill once in your project and answer the five setup questions to generate the model preferences file.
