What problem does it solve? Baseline metrics exported from one repository are often reused as fallback figures in another project, silently hiding their foreign origin and masking thin or missing data behind confident-looking totals. This Skill enforces honest reporting rules whenever you author, update, or render such a JSON baseline-data artifact. ## Core Features & Use Cases - Foreign-Origin Disclosure: Every table, chart caption, CLI print, or report section that surfaces the data states plainly that the figures come from a different repository than the one being reported on. - Exact Rollup Accounting: Aggregates use the precise formula tokens_billable(r) = tokens.input + tokens.output + tokens.cache_creation, always excluding cache_read, and treat null token records as exact zero contribution with no interpolation or backfilling. - Sample-Size Transparency: Every aggregate is paired with its contributing record count and the count of records excluded because capture_method was "unavailable", reported per tier or segment rather than collapsed into one repo-wide total. - Use Case: When generating a baseline.json of LLM token costs from one project to serve as fallback estimates in another, this Skill ensures consumers see where the data came from and how thin each tier's sample really is. ## Quick Start Use the baseline-data-honesty skill to update baseline.json so every aggregate shows its sample size, unavailable-record count, and foreign-repository origin.
