Scripts and agents

stage --async saves an interview and prints its first question batch. The batch contains a JSON Schema for the questions available now. A script or agent sends a JSON answers document keyed by question id. continue returns either the next batch or a completed result. apply writes files only after the interview is complete.

toha stage ./docs/examples/demo ./notes --async
echo '{"title":"Sample note","topic":"Research"}' | toha continue ./notes -
toha apply ./notes

continue on a complete interview prints the complete result again and exits 0. apply <template> <target> with an interview of that template staged at the target applies it as apply <target> does. A refused command names the commands that fit on standard error.

stage --async exits 4 when it emits a batch. continue exits 4 when questions remain or answers are rejected, and 0 when complete. apply exits 0 after writing, 3 when hooks need trust, 4 for an incomplete interview without a terminal to prompt in, and 1 for a conflict or other failure. An ambiguous installed template name exits 5. See the CLI specification for all exit codes.

For a headless run, pass a JSON answers file to apply:

printf '%s\n' '{"title":"Sample note","topic":"Research"}' > answers.json
toha apply ./docs/examples/demo ./notes --answers answers.json

A headless run with a missing required answer stages the interview and emits a batch; with --dry-run it emits the batch and stages nothing. Answers for later questions can be supplied early; unknown ids are errors. The interview protocol defines batch, result, and answers documents.

The binary carries agent guidance. Use toha skills list to discover it, toha skills view <name> to read one skill, or toha skills view <name> --export <path> to export it for an agent.