What problem does it solve? AI assistants do not remember between sessions, and large projects do not fit in one. This Skill gives a project a durable, resumable task state: where each task stands, what it produced, and what comes next, so any session can pick up exactly where the last one stopped. ## Core Features & Use Cases - Single source of truth: an estado.json state machine (one entry per task with status, priority, artifacts, and optional blocking edges) from which a human-readable TAREFAS.md view is deterministically regenerated. - Controlled status transitions: a canonical 9-status enum (nao_iniciada through concluida/abandonada) with validated transitions, safe writes via backup plus atomic temp-file rename, and concurrent-write detection through atualizado_em checks. - Orchestration support: stores round counters, scoreboards with per-lens granularity, pinned success predicates with fail-closed revalidation, and write-once execution receipts for iterative loops. - Use Case: You return to a project after several days and ask where things stopped. The Skill reads estado.json, reports each task's status, and resumes work without reconstructing context. ## Quick Start Ask the assistant to read the project state and tell you where each task stopped and what to do next.
