Do I need to write anything before running this?
No. A rambling walkthrough is the expected input. The pack does the structuring, and it asks follow-up questions where your description was ambiguous.
Ops
Capture how a recurring task actually gets done before the knowledge lives in exactly one head.
Follows the Agent Skills open format, works with Claude Code, Cursor, Hermes Agent, OpenClaw, and 30+ compatible agents.
What it does
This skill turns the way you already explain a process out loud into a written SOP. You talk or type through how the work gets done, and it produces numbered steps with an owner, a cadence, a definition of done for each step, and the decision points where people usually get stuck.
It is built for the moment before a hire or a handoff, when the process exists but only in one person's habits. Rather than asking you to write documentation, it interrogates the walkthrough: what triggers this, what does finished look like, what happens when the input is missing.
The output is written so either a teammate or an agent can execute it. Steps are specific, decision rules are explicit, and anything you could not answer during the walkthrough is listed as an open question instead of being quietly guessed at.
Who it is for
Requirements
Process walkthrough
A rough transcript, notes, or spoken run-through of how the task gets done
Owner and cadence
Who runs this today and how often it happens
Tools involved
Optional. The systems each step touches, so steps name the right place.
Docs or wiki
Optional. Where the finished SOP lives so the team can find it.
Sample output
SOP: Monthly client reporting - Northwind Ceramics account Owner: Account lead Trigger: First Tuesday of the month Estimated time: 40 minutes 1. Pull last month's numbers into the reporting sheet. Done when every KPI row has a value and a prior-month comparison. 2. Flag any metric that moved more than 15 percent and write one line on the likely cause. Done when each flag has a cause or an explicit 'unknown'. 3. Draft the client summary in the standard four-section format. Done when the draft names one recommended action. 4. Send to the account lead for review by 3pm. Do not send to the client without review. Decision rules: - If a data source is missing, ship the report with the gap labeled rather than delaying the send. - If a metric moved more than 40 percent, call the client before the report goes out. Open questions: Who runs this when the account lead is on leave?
What's in this pack
Included: Outcome definition
What exact job this pack solves
Included: Operating instructions
Step-by-step workflow the agent follows
Included: Input contract
Required and optional fields defined as a schema
Included: Output contract
Exact output structure the agent must return
Included: Worked examples(6 files)
3 standard, 1 ambiguous, 1 bad-fit, and 1 edge-case scenario
Test & eval set
FAQ
No. A rambling walkthrough is the expected input. The pack does the structuring, and it asks follow-up questions where your description was ambiguous.
Then the variation is the useful part. The pack captures the stable steps and turns the rest into explicit decision rules, so exceptions stop being improvised.
Yes. Steps are written with concrete triggers and a definition of done, which is what an agent needs to execute rather than paraphrase.
Bundle option
This pack is also included in Founder Ops Pack, which groups related workflows for a faster setup.
View bundleRelated packs
Ops
Turn a signed deal into a week-one onboarding plan, kickoff email, and task checklist.
Test cases with expected outputs and pass/fail criteria
Included: QA checklist
Human review checklist for output quality
Ops
Summarize weekly metrics, outliers, likely drivers, and recommended follow-up in one operating brief.
Support
Classify support tickets, flag urgency, draft the next action, and reduce queue chaos for lean teams.