← Back to library

Turn one finished chore into a saved agent skill

Argona0x walks through training a desktop worker the way xAI's manual describes: describe the chore like onboarding instructions for a new hire, try it once on something safe, store the method with an approval rule, then schedule weekday runs that report instead of inventing missing data. Their example was filing a receipts folder.

  • OutcomeBrief an agent on a real job you just did, correct its first run, save the method as a skill with an approval rule, then schedule it only after it works.
  • SetupMinutes
  • LevelStarter
  • CreditArgona0x (@Argona0x)
View original
Works with
Grok BotClaudeChatGPT agent

Published Oct 23, 2026Updated Oct 23, 2026AI dry run on sample input: passed Oct 23, 2026Source verified Oct 23, 2026

When should you use it?

  • You repeat the same folder chore weekly
  • You want a schedule only after a checked method exists
  • You can name which steps always require your OK

Not for: Scheduling a skill nobody reviewed; Letting it invent yesterday's numbers when a source file is missing

What do you need, and what do you get?

You need

  • One real job described like a new-hire brief
  • A safe sample folder
  • The approval rule (card numbers, for example)

You get

  • A saved skill
  • A first-run report
  • A schedule that fails closed when inputs are missing

How does it work, step by step?

  1. Write the briefSay what finished looks like, which files to use, what finished deliverable to return, and where it must ask you.
  2. Run once safelyEdit the output until you would put your name on it.
  3. Save the skillStore the method plus the one approval rule.
  4. Schedule with a fail-closed ruleIf inputs are absent, tell you instead of using old numbers.

The prompt

Our own prompt, inspired by Argona0x (@Argona0x) and written to work in Grok Bot, Claude and ChatGPT agent. Values in lime are placeholders.

Fill in the blanks (optional; the copied prompt uses your values)

Prompt5 variables
Job brief (onboarding style): {brief}.
Run once on {sample_path}. Show me every item you are unsure about.
After I correct you, save this as skill {skill_name} with approval rule: {approval_rule}.
Then propose a schedule: {cadence}. If the source path or files are gone, message me instead of reusing prior outputs.

Inspired by Argona0x (@Argona0x): Train a worker on one job then save it as a skill. The prompt above is written by playbooks.bot in our own words; nothing from the original is quoted.

What can go wrong?

  • A schedule on top of an unchecked method just automates mistakes.
  • Do not skip the approval rule for money or identity fields.

Our take

The valuable part of this post is the order: brief, correct, save, then schedule. Too many people jump to a cron job. We keep the paperwork-folder example and strip the unrelated deal-size rhetoric. In our dry run on a sample paperwork folder, the assistant sorted fourteen items, parked two pending a card number, and refused to run the schedule when the folder path was missing. Name the skill after the job, not the tool. Rehearse the fail-closed path by temporarily moving the folder. Keep approval rules next to the skill so successors inherit them. Do not schedule until a human has signed off on two clean runs. Money and identity fields stay behind approvals forever. Name the skill after the job, not the tool. Rehearse the fail-closed path by temporarily moving the folder. Keep approval rules next to the skill so successors inherit them. Do not schedule until a human has signed off on two clean runs. Money and identity fields stay behind approvals forever.

Where the time goes back: You pay the training cost once, then weekday runs start from a method you already corrected.

FAQ

Is this only for receipts?
Receipts are the example. The same four messages work for any repeatable folder chore.
Which model does it use?
Follow your product's rules. The playbook is about the training order, not picking a model.

Source and credit

Inspired by Argona0x (@Argona0x). We describe the idea in our own words, link to the original, and wrote the prompt ourselves; nothing is quoted.

Inspired by
Train a worker on one job then save it as a skill
Creator
Argona0x (@Argona0x)
Published on
X
Original date
Aug 19, 2026
License
Not openly licensed (the poster keeps all rights), so we quote nothing and link to the original
Originally for
Grok Bot
AI dry run
Oct 23, 2026: passed. Dry-run tested by our AI test agent on sample input. PASS: Fail-closed schedule; approval rule.
AI dry run on sample data; not run with live connected accounts. The playbooks.bot test agent filled the prompt with synthetic inputs, produced the output and checked it against the page's stated outputs and guardrails.
Verified
Oct 23, 2026: source link resolves to the original, the credited creator published it, our excerpt or prompt matches it, and the license allows our use.

Are you the creator? Claim or correct this page.

More in Operations & Admin

All 203 Operations & Admin playbooks →

Related playbooks

Operations & AdminSkill✓ AI dry run

Have Claude fill PDF forms, even ones with no fields

Give Claude the pdf skill so that when you hand it a form and the answers, it first checks whether the file has real form fields, fills those by their IDs if it does, places text by measured coordinates on flat or scanned forms if it does not, validates every box before writing, and renders the filled pages as images so you can see where each answer landed before you sign or send it.

Credit: Anthropic

Minutes
Operations & AdminWorkflow✓ AI dry run

Log a sales order after checking its paperwork matches

Give one spoken instruction and get the order logged, cross-checked against its paperwork, the commission tracked and a client email drafted.

Credit: Elvis Rigglet (@rassailor)

An afternoon+
Operations & AdminRoutine✓ AI dry run

Get a weekly performance report on your rental listings

Have an agent review your short-term rental pricing and bookings every week and send a short report with suggested price changes you approve.

Credit: madaxe_again on Hacker News

Under an hour
Operations & AdminSkill✓ AI dry run

Send suppliers an RFQ package that makes bids comparable

Use the supplier-quote-request skill from Anthropic's Claude for Manufacturing plugin to assemble a request for quotes with the drawing at its confirmed revision, a sourced demand picture, quality clauses, commercial terms and a required bid layout, list what is missing, draft one email per supplier for the buyer to send, then read the returned bids into one shape and compare them by quantity break.

Credit: anthropics

Under an hour
Operations & AdminWorkflow✓ AI dry run

Forward an email and let an assistant book the meeting

Forward a meeting request to a labeled inbox and have an AI assistant coordinate a time with the other person and add a video link.

Credit: Eddie Soehnel (@edsoehnel)

Under an hour