Owner Actions — Skill Shareware×
Address
https://johnarndt.net/skills/owner-actions

S-04 · Soxoa Ops · Run the business · All 26 skills

The list your AI can't finish for you

Agents can run commands, edit code and check results. They shouldn't pay, sign in as you, send to customers or make the call. A dated owner-action list draws that line every time. Free Claude skill.

Where it came from: My running John Actions files: what the agent finished and verified versus the clicks only I can make.

The usual way AI work goes wrong for a business owner isn't a dramatic failure. It's quieter than that. The agent finishes, reports success, and three things are still waiting on the owner. Nobody wrote them down. The payment webhook isn't on. The new sender domain isn't verified. A decision about the free tier is sitting in paragraph nine of a long reply. A week later something is broken, and it turns out it was never set up.

owner-actions exists so that every piece of agent work ends with one clear handoff.

Read the full article on soxoa.com

Install
/plugin marketplace add johnarndt/soxoa-claude-plugins
/plugin install soxoa-ops@soxoa-plugins
Download owner-actions.zipView on GitHub
View SKILL.md
---
name: owner-actions
description: "Use at the end of agent work, or when asked, to separate what the AI finished and verified from what only the owner can do. Triggers include 'what do I need to do', 'what's left for me', 'make me a to-do list from this', 'what's blocked on me', 'list the dashboard steps', or the end of a long session with logins, payments, approvals or account settings involved. Produces a dated, copy-paste-ready owner action list ordered by urgency, with proof for everything marked done."
---

# Owner actions

Agents can run commands, edit code and check results. They should not click "pay", sign in as the owner, grant access, approve a message to a customer, or make the strategy call. This skill draws that line in writing, every time, so nothing waits silently on a person who does not know it is waiting.

## What belongs to the owner

- Signing in, OAuth consent, creating accounts, multi-factor prompts.
- Payments, plan changes, refunds, anything that spends money.
- Sending to customers, prospects or the public: emails, posts, ads, store submissions.
- Account, security, DNS and billing settings in third-party dashboards.
- Legal, compliance or contractual sign-off.
- Decisions: what to build, what to stop, what to charge.
- Rotating any credential that appeared in chat or logs.

Everything else (running commands, editing files, testing, drafting, verifying) the agent does itself and records as done.

## Output

Write `OWNER_ACTIONS_YYYY-MM-DD.md` (or the name the user prefers) and summarize it in chat.

```markdown
# Owner actions · 2026-10-25

## Done (verified)
1. ~~Fixed the signup alert on Acme~~ · test signup at 10:42 produced one email · commit 3f9c2e1

## Do now (money or customers affected)
2. **Turn on the payment webhook.** Dashboard → Developers → Webhooks → Add endpoint → paste `https://acme.example/api/webhook` → select `checkout.session.completed`. Why: paid orders are not being recorded. If skipped: every sale needs manual entry.

## Toggles (flip when ready)
3. …

## Decisions
4. **Keep or stop the free tier?** Options and what each changes, in two lines.
```

Each item has: the exact place (URL or menu path), the exact value or command to paste, one line on why, and what happens if it is not done. Number items across sections so the owner can reply "did 2 and 4".

## Rules

- Never write a secret, password or full key into the list. Name it ("the payment provider's webhook signing secret") and say where it goes.
- "Done" requires proof: a command output, a test result, a live URL, a commit. Without proof it stays in a "Done, not verified" line.
- Do the work before listing it. If the agent can do a step, it does the step instead of listing it.
- Keep it short. If the owner has more than about ten open items, put the top five first and the rest under "Later".
- Carry unfinished items forward into the next dated file; do not silently drop them.
- A gate that exists on purpose (a human review queue, an approval step) is listed as a gate, not reported as a bug.