What problem does it solve? Engineering teams using AI coding agents have no reliable record of which work items consumed which effort and tokens. This Skill captures work as it happens — opening tasks with pre-registered estimates, closing them with shipped evidence, and pinning session token spend to tracker keys — so later reports rest on receipts rather than memory. ## Core Features & Use Cases - Pre-registered task opening: Records a without-Claude time estimate sealed in SHA-256 escrow before work begins, making later time-saved claims credible. - Closing shipped work: Appends a shipped entry superseding the opened one, with PR URLs, metered tokens, and computed time saved at facts or judgment tier. - Session pinning and mining: Binds a session's token spend to a tracker key at confidence 1.0, retracts pins via correction events, and mines pending session captures deterministically. - Use Case: A developer says "I'm starting PLAT-482, log it", gives a 10–16h estimate, ships the PR two days later, and closes the entry with actual hours and metered tokens for sprint reporting. ## Quick Start Tell the assistant "log this to my waybill ledger" or "I'm starting PLAT-123, log it" to record the work item in your local ledger.