Transcript in. Decisions, tasks, owners and a follow-up email out.

---
name: client-calls
description: Turn a client call transcript into decisions, tasks with owners, and a follow-up email. Use whenever I paste a call transcript or notes from a meeting.
---

# Client calls

When I give you a transcript:

1. Decisions. What was actually settled. Only things that were decided, not things that were discussed.
2. Tasks. One line each, with an owner (me or the client) and a due date if one was said out loud.
3. Open questions. Anything left hanging that has to be answered before work can start.
4. Follow-up email. Short and warm, written to the client. Lead with what we decided, then their tasks, then mine. No preamble, no recap of the small talk.

If an owner or a date was never said, write "not stated". Do not guess one.

What it does. Turns a forty-five minute call into the four things you actually needed from it, and writes the follow-up while the call is still fresh. The "not stated" rule is doing quiet work — without it you get invented deadlines that feel authoritative enough to end up in your calendar.

What it does not do. It cannot fetch the transcript. You still paste it in, or connect a recorder yourself. And it only knows what was said out loud — the thing you agreed to in your head on the call is not in the file, so it will not appear in the tasks.

One long video into five angles, five Reel scripts and a filming list.

---
name: content-engine
description: Turn one long video transcript into five short-form angles, Reel scripts and a filming list. Use when I paste a long transcript or a long-form script.
---

# Content engine

From one long transcript, give me:

1. Five angles. Five different arguments that are genuinely in this video. Five whole ideas, not five ways of phrasing the same one.
2. A Reel script per angle. Hook in the first line, then the point, then one specific example from the transcript. Spoken language, no headings, under 45 seconds read aloud.
3. A filming list. What I actually have to get on camera, grouped so I can shoot the lot in one sitting.

Use my own words from the transcript wherever they are better than yours. Do not invent examples I did not give you.

What it does. Gets five distinct pieces out of something you already recorded, then tells you what to film in one batch instead of five separate setups. The "do not invent examples" line is the difference between your stories and generic ones with your name on them.

What it does not do. It will not tell you which angle will perform — it has no idea what your audience responds to. And five is a target, not a promise: a thin transcript gives you two strong angles and three stretches. Cut the stretches rather than filming them.

Reads your voice guide and quotes real examples before it writes a word.

---
name: brand-brain
description: Read my voice guide and quote real examples from it before writing anything as me. Use whenever I ask for copy, captions, emails or posts written in my voice.
---

# Brand brain

Before you write a single line as me:

1. Read my voice guide and my sample posts.
2. Quote three real lines from the samples that match the job in front of you, and say in one sentence why each one works.
3. Then write, matching the rhythm of those three lines.

Rules:
- Australian English.
- Short punchy sentences, then a longer one for breath. Never three tidy items in a row.
- If you cannot find a real example that matches this job, say so and ask me for one. Do not guess at the voice.

What it does. The quoting step is the whole trick. Made to find and quote three real lines first, it has to actually read your writing instead of generating its idea of "your voice" from your industry. The output changes noticeably, and you can see why it changed.

What it does not do. It has no voice guide unless you give it one. A skill is instructions, not a folder of your writing — so you either paste the samples in, or point it at a file it can read. Skip that and step 2 becomes it quoting itself, which is worse than not having the skill.

Orders against payments. Mismatches flagged, subscriptions listed.

---
name: money-check
description: Compare an orders export against a payments export, flag mismatches and list active subscriptions. Use when I give you sales and payment data to reconcile.
---

# Money check

Given an orders list and a payments list:

1. Match them on amount, date and customer.
2. Flag every mismatch. An order with no payment, a payment with no order, or an amount that does not line up. One line each, showing both numbers.
3. List active subscriptions. What recurs, how much, how often, and when it next charges.
4. Tell me what you could not match and why, BEFORE you give me any totals.

Never round. Never fill a gap with an estimate. If a field is missing from the data, say the field is missing.

What it does. Catches the order that never got paid and the subscription you forgot you were still on, in a couple of minutes instead of an afternoon of squinting at two spreadsheets. Putting the unmatched items before the totals stops a tidy-looking number hiding a hole.

What it does not do. This is not bookkeeping and it is definitely not tax advice. It is a mismatch finder across two lists you handed it. It knows nothing about refunds, chargebacks, or anything that never made it into the export — so check every flag against the real account before you act on it or chase anyone.

Reads your calendar, tasks and email, hands back three decisions.

---
name: monday-ceo
description: Read my calendar, tasks and inbox and return the three decisions only I can make this week. Use on a Monday, or whenever I ask what actually needs me.
---

# Monday CEO

Read my calendar for the week ahead, my task list and my unread email. Then give me:

1. Three decisions only I can make. Not tasks. Decisions, where someone is waiting on my answer and nobody else can give it.
2. What is genuinely time-sensitive this week, with the date it stops being possible.
3. What I should hand off or delete, and who to.

Three decisions maximum. If there are more than three, pick the three with the biggest consequences and tell me what you left out.

What it does. Separates the three things that actually need you from the pile that only feels like it does. The hard cap is the useful part — a list of eleven priorities is not a list of priorities, and it will happily give you eleven if you let it.

What it does not do. It cannot see your calendar or your inbox on its own. You have to connect them — a connector on claude.ai, or a tool in Claude Code — or paste the week in by hand. Without that it will produce three confident, entirely fictional decisions, which is the worst possible version of this skill.

The follow-up email is the bit that costs you clients — not because you never send it, but because you send it Thursday and it has lost its edges. This one has the highest ratio of relief to effort of the five, which is why it is first.

You have hours of good material and no time to sit and pull it apart. Run one existing video through and you get next week's filming list off something you already recorded. Batch what it hands you.

Build this one before any of the writing ones and every piece of copy after it gets better. Worth doing properly — feed it your best twenty posts, not three you found quickly, because it can only quote what you gave it.

Once a month, not once a week. Most people find the unpaid order on the first run and a subscription they thought they cancelled on the second. Check anything it flags against the real account before you chase it.

Build this one last, because it needs your calendar and inbox connected to be worth anything. Once they are, it is the one you will run most — nine minutes on a Monday and you know the three things that genuinely need you.