test-a-generator-for-what-it-invents

Check a document generator against sparse and missing input, where fabrication is disqualifying rather than one score among several.

Procedure

  1. start start. Goes next to s1
  2. action Generate from complete input, RW (reversible-write). Goes on_fail to s9; next to s2
  3. action Generate from the sparse case, RW (reversible-write). Goes on_fail to s9; next to s3
  4. action Generate with the number left empty, RW (reversible-write). Goes on_fail to s9; next to s4
  5. decision Did it invent anything?. Goes if fabricated to s9; if default to s5
  6. action Generate the same thing three times, RW (reversible-write). Goes on_fail to s9; next to s6
  7. check The longest expected case completes. Goes on_fail to s9; next to s7
  8. check It writes a new file in a governed place. Goes on_fail to s9; next to s8
  9. decision Is this output format covered by labelling?. Goes if not_covered to s10; if default to s11
  10. action Constrain the format, or record the gap, RW (reversible-write), requires approval. Goes on_fail to s9; next to __end_success
  11. action Record what was tested, and against what, RO (read-only). Goes next to __end_success
  12. escalate Do not launch it. Terminal
  13. end success. Terminal

Open the graph on its own page — larger, printable, and an address you can send.

Walk it without running it — you answer what your environment would answer, and the path, the capabilities, the approvals and the distance back to an undo come out of the document. Nothing executes; there is no Run button here and there will not be one.

Trust

T2 · attested 2026-10-11 · runbooks.directory

Nobody has attested this record. That is the ordinary case: T2 is what the catalog can check on its own, and T3 needs somebody to run the procedure and sign for it.

  • Link liveness has not been checked: it needs a network and the catalog builds without one. Run pnpm gates --online where there is one.

Steps

  1. s1 action — Generate from complete input (reversible-write)
  2. s2 action — Generate from the sparse case (reversible-write)
  3. s3 action — Generate with the number left empty (reversible-write)
  4. s4 decision — Did it invent anything?
  5. s5 action — Generate the same thing three times (reversible-write)
  6. s6 check — The longest expected case completes
  7. s7 check — It writes a new file in a governed place
  8. s8 decision — Is this output format covered by labelling?
  9. s10 action — Constrain the format, or record the gap (reversible-write) — requires approval
  10. s11 action — Record what was tested, and against what (read-only)
  11. s9 escalate — Do not launch it

Capabilities

This is a privilege request. Check it against your own allowlist before running.

    Provenance

    Source
    https://github.com/microsoft/ai-agent-runbooks/blob/083f33eec889/02-patterns/Branded-Office-Artifact-Generation/Branded-Office-Artifact-Generation-Runbook.md
    Fetched
    2026-09-04
    Source last checked
    not since it was captured on 2026-09-04 — recorded as current
    Licence
    MIT
    How it got here
    Rewritten here from the source above, under its licence. The words and the steps are this catalog's; the procedure it describes is not this catalog's invention, and the source is where to check it.
    Highest risk in this procedure
    reversible-write

    Take it with you

    Run it under supervision

    Runs here, on your machine, under supervision: every step is proposed before it happens, an approval gate stops for you, and a failed expectation goes where the procedure says rather than on to the next step. Nothing is sent anywhere. To supervise an agent that is doing the work instead, runbooks hook sits in its tool-call path.

    npx @runbooks/cli run adapted/test-a-generator-for-what-it-invents@1.0.0 --as "$(whoami)" \
  --input sparse_case=<a real input with most fields genuinely missing> \
  --input numeric_field=<a number the output normally carries, left empty on purpose>