SOP / Runbook
Turns rough process notes into a runbook with an owner, trigger, checked steps, failure handling and an escalation path.
When to use itUse it when a process lives in someone's head and keeps breaking at handover. It documents the process as your notes describe it and flags missing owners or contacts; it cannot tell you whether the process itself is sound.
- Type
- Playbook
- Inputs it asks for
- 07
- Sections it returns
- 09
- Quality checks
- 08
On this page5 sections
Asks for
What you hand it before it runs. Required ones are marked.
- 01Process notesLong textRequired
How the process runs today: notes, chat threads, a walkthrough transcript, an existing doc, the commands and screens used. Rough is fine. Only what is here gets documented.
- 02Tested againstText
Systems and versions the steps were last run on, and the date, e.g. billing admin v3, run on 2026-10-08. Left blank, the runbook flags it.
- 03OwnerText
The role or person accountable for the process. Left blank, it is flagged, never guessed.
- 04TriggerText
The event, alert or schedule that starts the process, e.g. a disk-usage alert, or the first working day of the month.
- 05Who runs itText
The role and experience of the person following it, e.g. an on-call engineer on their first rota.
- 06Escalation contactsLong text
Who to contact when a step fails, in order, how to reach them (role, channel or rota) and after how long. Never invented.
- 07RiskSelect
Customer-facing and irreversible processes get approval checkpoints and warnings before the risky steps.
Method
The instruction the model follows, word for word as it ships in the SKILL.md.
You write a standard operating procedure or runbook from the Process notes. The Technical Documentation Writer rules apply throughout (source only, [VERIFY] flags and the To verify list, show then explain, action plus expected result, ALL_CAPS placeholders, labelled shell syntax, precise verbs). The precise-verb rule matters most here: every step names the concrete action, the system or screen, and the thing acted on. This method adds only what is specific to SOPs.
The runbook succeeds when the person in Who runs it can complete the process without the author present, can tell at each checkpoint whether it is going right, and knows whom to contact when it is not.
Structure:
- H1: the process name. Directly under it, one line: "Tested against: " followed by the Tested against value. If Tested against is blank, write "Tested against:
[VERIFY: version and date last run]". - Purpose: one or two sentences on what the process achieves and, where the notes say, what goes wrong if it is skipped or late.
- Owner: the accountable role from Owner or the notes. If neither names one, write "
[VERIFY: owner]". Never assign an owner yourself. - Trigger: the event, alert or schedule that starts it, from Trigger or the notes, and any deadline the notes give.
- Before you start: access, tools, permissions and information needed, as bullets.
- Procedure: numbered steps, one action each. Start the step with the actor when more than one role takes part. Give the action with the command, screen path or field where the notes show one, then an Expected result line. Write decision points as "If X, go to step N. Otherwise, continue." When Risk is Customer-facing or Incident or irreversible changes, put an approval checkpoint (the approver from the notes, or
[VERIFY]) and a > **Warning:** saying what cannot be reversed before each step that cannot be undone. Add no steps the notes do not imply; where a step is plainly missing between two others, insert a placeholder step flagged with what appears to be missing. - Quality checks: the checks that confirm the whole process succeeded, each one observable (a record, a report value, a message sent).
- Failure handling: a table with the columns Step, Symptom, Action, Stop or continue, for the failures the notes describe, including how to roll back where the notes say. Flag any irreversible step that has no failure path in the notes.
- Escalation path: an ordered list of whom to contact, how, and after how long or under what condition, from Escalation contacts or the notes only. If none are supplied, write "
[VERIFY: escalation contacts]". Never invent names, channels or rotas. - To verify: the numbered list of every
[VERIFY]flag, each with its section or step and what would settle it.
Where the notes describe the same step two ways (different people doing it differently), document the version the most recent or most specific note supports and list the conflict under To verify. The runbook records the process as described; it cannot show that the process is sound.
The shape of one procedure step (shape only; the content must come from the notes):
Example
Length: as short as the process allows, with steps of one or two sentences plus any command. If the notes do not show the sequence of the process, ask up to three specific questions instead of writing.
Returns
Every reply comes back in this shape, in this order.
- 01Purpose
- 02Owner
- 03Trigger
- 04Before you start
- 05Procedure
- 06Quality checks
- 07Failure handling
- 08Escalation path
- 09To verify
Quality checks
Read the reply against these before you use it. Tick them off as you go; nothing is saved.
0 of 8 checked
SKILL.md
The whole playbook as one skill file, for Claude, Codex or any tool that reads SKILL.md.
---name: "sop-runbook"description: "Turns rough process notes into a runbook with an owner, trigger, checked steps, failure handling and an escalation path."--- # SOP / Runbook You write a standard operating procedure or runbook from the Process notes. The Technical Documentation Writer rules apply throughout (source only, [VERIFY] flags and the To verify list, show then explain, action plus expected result, ALL_CAPS placeholders, labelled shell syntax, precise verbs). The precise-verb rule matters most here: every step names the concrete action, the system or screen, and the thing acted on. This method adds only what is specific to SOPs. The runbook succeeds when the person in Who runs it can complete the process without the author present, can tell at each checkpoint whether it is going right, and knows whom to contact when it is not. Structure:- H1: the process name. Directly under it, one line: "Tested against: " followed by the Tested against value. If Tested against is blank, write "Tested against: [VERIFY: version and date last run]".- Purpose: one or two sentences on what the process achieves and, where the notes say, what goes wrong if it is skipped or late.- Owner: the accountable role from Owner or the notes. If neither names one, write "[VERIFY: owner]". Never assign an owner yourself.- Trigger: the event, alert or schedule that starts it, from Trigger or the notes, and any deadline the notes give.- Before you start: access, tools, permissions and information needed, as bullets.- Procedure: numbered steps, one action each. Start the step with the actor when more than one role takes part. Give the action with the command, screen path or field where the notes show one, then an Expected result line. Write decision points as "If X, go to step N. Otherwise, continue." When Risk is Customer-facing or Incident or irreversible changes, put an approval checkpoint (the approver from the notes, or [VERIFY]) and a > **Warning:** saying what cannot be reversed before each step that cannot be undone. Add no steps the notes do not imply; where a step is plainly missing between two others, insert a placeholder step flagged with what appears to be missing.Used together with
- 01Technical Documentation WriterWrites setup guides, API references, READMEs and runbooks from your source material, and flags anything it cannot confirm.Operator
- 02API Endpoint ReferenceTurns route code, schemas or captured requests into endpoint reference docs with parameter tables, full examples and error fixes.Playbook
- 03READMEWrites a README from your manifest, code and tests: what it does, a quick start of five steps or fewer, real usage examples and licence.Playbook
Free to use and adapt under CC BY 4.0. Credit MyPromptVerse.
Built in MyPromptVerse. Follow the build →