a-prompt-the-filter-refuses
Reword a request a safety filter declined, in a fixed order, and keep the pair — because the refusal is usually about how it was asked rather than what was asked.
Procedure
- start start. Goes next to s1
- decision Is the refusal about the subject or the asking?. Goes if genuinely_refused to s7; if default to s2
- action Ask for retrieval, not characterisation, RO (read-only). Goes on_fail to s3; next to s6
- action Split it into narrower questions, RO (read-only). Goes on_fail to s4; next to s6
- action Remove what asks it to infer or judge, RO (read-only). Goes on_fail to s5; next to s6
- decision Has anything worked?. Goes if none_worked to s7; if default to s6
- action Record the pair, blocked and working, RO (read-only). Goes on_fail to s7; next to __end_success
- escalate Escalate, with the three attempts attached. Terminal
- 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 --onlinewhere there is one.
Steps
s1decision — Is the refusal about the subject or the asking?s2action — Ask for retrieval, not characterisation (read-only)s3action — Split it into narrower questions (read-only)s4action — Remove what asks it to infer or judge (read-only)s5decision — Has anything worked?s6action — Record the pair, blocked and working (read-only)s7escalate — Escalate, with the three attempts attached
Capabilities
This is a privilege request. Check it against your own allowlist before running.
Provenance
- Source
- https://github.com/microsoft/ai-agent-runbooks/blob/083f33eec889/01-scenarios/Client-Meeting-Preparation-Agent/3.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
- read-only
Take it with you
- SKILL.md — the record as an Agent Skills file
- workflow.yaml — Open Workflow DSL 1.0.3, for a runner
- record.json — the catalog's own JSON
- graph.svg — the graph, standalone
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/a-prompt-the-filter-refuses@1.0.0 --as "$(whoami)" \
--input prompt=<the request that was refused, exactly as it was sent>