What problem does it solve? After a sprint or project ends, teams often skip structured retrospectives or run feeling-based reviews that produce no calibration value. This Skill enforces a data-driven retrospective SOP that compares estimated versus actual hours, classifies deviation causes, and converts findings into WBS estimation calibration factors so the next quote is more accurate. ## Core Features & Use Cases - Four-block retrospective workflow: numeric comparison of estimated vs actual hours and gross margin, Keep/Improve/Change analysis, WBS deviation classification (scope creep, technical underestimate, external dependency, optimism bias), and tracked action items. - Cumulative WBS calibration table: maintains per-work-package calibration multipliers (e.g., third-party API integration × 1.35) that feed back into future estimates. - Automatic project-state update: writes sprint results and health status into the TRACKER client's project-state.json and commits it via git and Cloudflare Pages deploy. - Use Case: After delivering a client project where actual hours exceeded the estimate by 35%, run the retrospective to classify the deviation as technical underestimation, raise the calibration multiplier for that work package type, and produce action items with owners and deadlines. ## Quick Start Ask the AI to run a retrospective for the just-finished sprint or client project, providing the estimated and actual hours per module so it can produce the calibration report and update the project state.
