TaskMaster - AI Cost Optimizer

**Quick judgment** TaskMaster is a solid, general-purpose AI project manager and cost optimizer that is especially useful for complex, multi-step workflows where you care about both model selection and spend control. It appears well thought-out for research, development, and operational task orchestration, but real-world robustness will depend on how accurately it measures costs and how reliably its delegation/aggregation logic behaves under load. **What it does well** - Breaks complex projects into smaller tasks and assigns each to a model tier (Haiku/Sonnet/Opus) based on complexity. - Spawns sub-agents with model constraints and can run them in parallel for faster turnaround. - Tracks token usage and total project budget, with limits and alerts to avoid overspending. - Provides progress tracking, basic retry logic, and final deliverable compilation. **Key risks / limitations** - **Cost accounting accuracy**: If underlying token or price tracking is off, you may think you are within budget when you are not. Use conservative budget margins until validated with your own billing data. - **Model-selection misfires**: Automatic triage may over- or under-allocate model capability (e.g., using a cheaper model for tasks that actually need deeper reasoning). Use FORCE overrides for critical tasks. - **Operational complexity**: Parallel sub-agents and isolated sessions increase orchestration complexity; subtle bugs could cause lost context, duplicated work, or inconsistent aggregation. - **No explicit domain safety controls**: The skill focuses on cost and workflow, not on content safety or domain-specific constraints. You must enforce your own data-handling and safety policies. **Recommended scenarios** - Multi-part research or analysis projects with clear budget ceilings and deadlines. - Building or iterating on software MVPs where different subtasks have clearly different complexity and cost profiles. - Teams wanting a repeatable pattern to delegate routine tasks to cheaper models while reserving expensive models for high-value reasoning. - Any user who frequently runs into token or cost overruns and needs more structured, budget-aware orchestration of AI work.