What problem does it solve? AI-assisted delivery flows consume tokens across models, stages, and sub-agents, and without a periodic review those costs accumulate unnoticed. This Skill reads the token usage already recorded by the bound delivery surface, identifies the top spending drivers, and turns them into concrete, prioritized cost-reduction actions. ## Core Features & Use Cases - Cost Data Aggregation: Collects flow runs in a configurable time window (default last 7 days) and aggregates total tokens, uncached tokens, per-model, per-stage, and sub-agent breakdowns. - Pattern Analysis: Detects the most expensive model and stage, compaction or truncation events, repeated prompts that could be cached, and runs restarted instead of handed off. - Prioritized Report: Outputs a structured weekly report with summary metrics, top cost-reduction tips, and repository-specific action items, with an optional full breakdown format. - Use Case: Run every Monday to review the previous week's flow runs, discover that a high-cost model handles low-complexity tasks, and get a checklist recommending a cheaper model tier for those agent files. ## Quick Start Run the weekly cost analysis for the last 7 days and show me the top cost-reduction tips from our recorded flow runs.
