factory-decisionslisted

# Read, explain and record a decision Input: the records under `.agents/factory/decisions/`, the story and the stage file each one names, and what the human says. Output: at most one `## Answer` section per record, in the human's words, on their say-so. You write no plan, no test and no code, and you never write an answer that was not given to you. The record's contract is in `factory-run/reference/file-contracts.md`, *Decision records*; read it once per project. The state of a record is read off the file: no `## Answer` is **open**; one with `answer:`, `by:` and `at:` is **answered**; a gate-written `## Applied` is **applied**; an `## Answer` missing the name or the time is a **draft**, and a draft unblocks nothing. ## Do 1. **List what is waiting.** Run `bash .agents/factory/factory.sh decisions` (add `--story <id>` for one story; `python3 .agents/factory/story-gate.py --list-decisions` where there is no runner copy). It prints one line per record — id, state, story and stage, when it was asked, the question — open ones first. Show that, in that order. Where the gate is not installed, read the directory yourself and say that you did. 2. **Explain one from its files, not from memory.** Read the record: the question, the options, the evidence the stage cited, its recommendation. Read the story it names and the stage file that asked (`tasks/<story>/<stage file>`, ending in `## needs-human`). Say what is asked, why the stage could not decide it, what each