What problem does it solve? Estimating how long an agent-led project will take and what it will cost is usually guesswork based on person-hours or invented labor rates. This Skill produces empirical delivery estimates from a validated scope, keeping wall-clock time, calendar elapsed time, API-equivalent token cost, marginal cash, and quota capacity as separate, honest figures. ## Core Features & Use Cases - Four modes: Estimate (forecast a scoped project), Reconcile (compare a prior estimate with verified actuals), Calibrate (rebuild empirical cohorts and priors), and Audit (check data integrity, freshness, and pricing provenance). - Deterministic offline helper: Runs a packaged Python helper with validated JSON inputs for request, prior, pricing, and quota data, never making provider calls or scraping prices. - Delivery-estimate checkpoint: Embeds a compact estimate section into formal implementation designs or plans, with typed results like estimate_unavailable instead of fabricated numbers. - Use Case: Before finalizing an implementation plan, invoke the skill to get P50/P80/P95 ranges for agent wall-clock time, calendar elapsed time, and API-equivalent token cost, with confidence and evidence coverage stated. ## Quick Start Ask the agent to estimate this project's wall-clock time and API-equivalent token cost based on the scoped phases and routes in the implementation plan.
