task-intakelisted

# Task Intake Turn raw requests into task cards that another person can execute without reopening the original conversation. Keep semantic judgment separate from deterministic ordering: first normalize and validate each card, then rank only executable work. ## Required companions Before drafting task prose, load `writing-skill` and choose one output language for the whole card. After drafting, apply `humanizer` without changing facts, scope, roles, dates, field values, or acceptance criteria. If either skill is unavailable, use `references/language-and-writing.md` as the fallback and disclose that fallback. ## Procedure 1. Read the target board schema, active projects or work scopes, people, workflow states, and existing task candidates. Search for semantic duplicates in the same scope before writing. 2. Split the request into atomic tasks. One executable card produces one reviewable result. Use a container only when the outcome cannot be accepted as one result; containers never enter the execution queue. 3. Preserve intent. Extract the requested result, source, constraints, named people, deadline, and real dependencies. Do not invent adjacent work. Put consequential unknowns in clarification rather than guessing. 4. Assign fields using `references/field-contract.md`. Distinguish the accountable owner, the executor or executors, and the reviewer. Relate the card to the project, initiative, area, or other canonical work scope supported by the board. 5. Write the title and