A Claude skill that reads your chats and your notes, names the choke point capping your growth with receipts, and designs the AI-powered fix you can ship this week.
← All Claude SkillsEvery business has one place where work piles up and waits, and it is usually waiting on a person. The Bottleneck Buster finds that choke point by reading what you have already written, your past chats and your notes, then scores every candidate and designs a concrete fix. Copy the two blocks below into Claude and you own it in under two minutes.
Step 1. Copy the skill block below.
Step 2. Open Claude and paste this instruction first, then paste the block underneath it:
Turn the block below into a Claude skill and install it. Create the skill file exactly as written. Then personalize it with smart defaults from what you know about me, my business, and where my work lives. Confirm when it is installed and tell me how to run it.
--- name: bottleneck-buster description: Identify the number one bottleneck in your business and bust it by leveraging AI. The skill finds bottlenecks by reading what already exists, your past chats and your second brain, then scores the candidates and designs the fix. Use this skill whenever the user says "run the bottleneck buster," "bottleneck buster," "find my bottleneck," "bust my bottleneck," "where is my business stuck," "what is slowing me down," or any variation of wanting to find where their business piles up and waits on a person, then design the AI-powered fix. Also trigger when the user describes a task they keep doing manually, a person their whole operation depends on, or a process that caps their growth. This is the skill that finds the choke point and designs the bust. --- # Bottleneck Buster Find the choke point. Design the bust. Point at the future where bottlenecks get found for you. Every business has one place where everything piles up and waits. Usually it waits on a person. Often that person is you. This skill identifies your number one bottleneck by design, runs it through a proven filter, and produces a concrete AI-powered plan to bust it this week. The skill is built to run on evidence, our past chats and your second brain, so identification happens without you doing homework. A skill that needs you to feed it everything just made you the bottleneck again. The machine brings you the suspects. You make the call. That call is your 8%. This skill has three layers. Run as deep as the moment calls for. Layer 1 alone is a win. All three is a transformation. ## Trigger Phrases - "run the bottleneck buster" - "bottleneck buster" - "find my bottleneck" - "bust my bottleneck" - "where is my business stuck" - Any description of a task done manually more than once, a person the operation depends on, or a process that caps growth ## Input One of the following: - Nothing at all. The default is the Evidence Scan: read our past chats and my second brain (Notion, Google Docs, wherever my notes live, ask me once and remember it) and come back with candidates. - A conversation where I describe my business, my week, or my frustration - A transcript of a team call or coaching call - A request for the Interview if I want to think out loud --- ## LAYER 1: IDENTIFY (Find the Choke Point) Three modes. Evidence Scan is the default because the best identification requires zero homework from me. Use the Interview when I want a conversation. Use Hybrid when I want both. ### Mode A: Evidence Scan (default, zero effort) The bottlenecks are already documented. I have been describing them for weeks in my chats and my notes. Read the record: 1. **Read our past chats.** Search past conversations for bottleneck signals: the same task described more than once, "I keep having to" language, work waiting on one named person, builds that stalled, the phrase "when I get time." 2. **Read my second brain.** Scan recent call transcripts, task lists, and meeting notes for pile-up signals: recurring action items that never close, one person's name attached to everything, a number that stopped moving. 3. **Surface 3 to 5 candidate bottlenecks with receipts.** Every candidate comes with evidence: where it showed up, how many times, and a direct quote when one exists. A bottleneck that appears in four separate conversations is announcing itself. ### Mode B: The Interview (the skill asks me) When evidence is thin or I want to think out loud, run the questions. Conversational, one question at a time, listening for the answer behind the answer. The bottleneck is rarely the first thing named. 1. **Where does work pile up and wait on one person?** Name the person. Name the pile. 2. **What do you do more than once that only you can do right now?** If you do something more than once, it can become a skill. If it can become a skill, it can run without you. 3. **What is the one thing that, if it flowed freely, would grow revenue the fastest?** This is your main thing. How much of the main thing is being eaten by everything else? 4. **Where does the founder become the ceiling?** What stops moving when you stop moving? 5. **What would the perfect world look like?** Describe it out loud. AI lives in a perfect world. The gap between today and that description is your bottleneck map. ### Mode C: Hybrid (scan, then confirm) Scan first, then ask only the questions the evidence could not answer. I confirm and add color rather than starting from a blank page. Fastest path to a high-confidence answer. ### The Bottleneck Score (not all bottlenecks are created equal) Every candidate gets scored on three dials, 1 to 10 each: - **Money:** How directly does busting this move revenue? - **Drag:** How much time and capacity does it eat every week? - **Bustability:** How much of the fix can AI own today? Total out of 30. Present the candidates as a ranked lineup. The machine brings the suspects. The owner makes the call. Highest score usually wins, and the owner can override, because judgment is the 8%. ### Layer 1 Output The ranked lineup, then the winner named in one sentence using this format: **The Bottleneck:** [Who or what] is stuck doing [low-leverage activity] instead of [the main thing], which caps [revenue, capacity, or growth outcome]. Example: Our closer is stuck qualifying leads over text instead of being on the phone with qualified buyers all day, which caps admissions. One bottleneck per run. The main thing stays the main thing. --- ## LAYER 2: BUST (Design the Fix) Run the named bottleneck through these four filters in order. Do not skip to building. ### Filter 1: The Big Dog Questions Answer all four before committing a single hour to a build: 1. **What is the clear definition of done?** Black and white. Data format. If you cannot describe done, you cannot delegate it to AI or anyone else. 2. **What is the asymmetric upside?** What does busting this unlock that compounds? 3. **What can AI and AI agents do right now at the click of a button?** Today's capability, honestly assessed. 4. **How is this connected to revenue?** Dollars per hour of the human currently doing it. Dollars gained when it flows. ### Filter 2: Buy or Build Ask: **does a tool already exist that does this?** Search before you build. Three weeks of CEO time at CEO rates building something a platform already does is the bottleneck reproducing itself. If the tool exists, your build is just training AI on how you use the tool, step by step. ### Filter 3: AI-First Rebuild Do not fix the old process with AI. Rebuild the process from zero assuming AI from the start. The 30-hour audit does not become a 27-hour audit with AI sprinkled in. It becomes a 3-hour audit designed AI-first. Ask: **if we were starting this today with AI at the center, what would we build?** Throw the old version in the trash and answer honestly. ### Filter 4: The One Piece Rule You do not need AI to own the whole process. Find the one piece in the chain that creates the most drag and bust that piece first. Qualification before conversation. Triage before treatment. Step one done, step two done, step three done. ### Layer 2 Output **The Bust Plan:** - The bottleneck (from Layer 1) - The definition of done - Buy or build decision, with the tool named if buy - The one piece to bust first - The first action, doable this week - Owner and deadline --- ## LAYER 3: THE 92/8 LEAP (Bottlenecks Found For You) This is where the skill becomes a system. In the 92/8 model, AI agents own 92% of execution and you own the 8% only you can do: judgment, creation, and coaching. The mature version of this skill runs as a routine without you. The Evidence Scan from Mode A becomes the Bottleneck Scout: 1. **The routine reviews your week on a schedule.** Transcripts, task lists, chats, and messages get scanned automatically for the same bottleneck signals Mode A hunts by hand. 2. **It surfaces one scored candidate per week** with the Layer 1 sentence already drafted and the receipts attached. 3. **You make the call.** Bust it, park it, or pass. Your judgment is the 8%. The identification was done for you. Every routine follows the standard: clear definition of done, always closes the loop by writing to your system of record (your tracker, wherever you track your work, ask me once and remember it), seven days of green lights before it earns more responsibility, and hard rules on what it will never do without you. The perfect world is a business where you wake up and the system tells you the number that matters. Everything between you and that number is a bottleneck waiting to be named. Now you have the machine that names them. ### Layer 3 Output One paragraph describing what the automated Bottleneck Scout routine would look like in this specific business: what it scans, when it runs, and what lands in front of the owner each week. --- ## Quality Standards - **One bottleneck per run.** Depth beats breadth. The main thing stays the main thing. - **Every bust plan ends in an action doable this week.** A plan without a first step is a wish. - **The definition of done is non-negotiable.** If done is fuzzy, go back to Filter 1 before building anything. - **Judgment-free zone.** The bottleneck is never a character flaw. It is a design opportunity. Audit without shame, then make one aligned move. - **Receipts over vibes.** In Evidence Scan mode, every candidate comes with proof of where it showed up. Evidence makes the pick easy and the conversation honest. - **This should feel like a treasure hunt.** Finding your bottleneck is finding money and time you already own. Keep the energy light. The skill does the digging, the owner gets the reveal. - **Bias toward the 60 to 80% bucket.** When choosing what to optimize, apply the 80/20 rule and build for the highest-probability outcome. ## Session Close End every run with three lines: 1. The bottleneck, named in one sentence 2. The first bust action and its deadline 3. The date of the next run. Bottlenecks are not a one-time find. The business will never not have one. Finding the next choke point is the system working.
Claude reads the block, builds the skill file, and installs it. From then on, you just run the skill in any conversation. No folders to find, no code to write.
Grow your business with AI, made simple and fun. Like having a Chief AI Officer by your side.
Here's the shape of what this skill hands you when it runs. Copy it, use it, keep moving.
You don't need to hold any of this in your head. That's the skill's job. If you're curious what it's doing for you, here's the idea underneath it.
Your bottlenecks are already documented. You have been describing them for weeks in your chats and your notes. This skill reads the record, scores what it finds, and hands you the one fix that moves revenue most.
Finds bottlenecks with receipts: where each one showed up, how many times, and a direct quote. The pick stays honest and the choice stays easy.
Ranks every candidate on Money, Drag, and Bustability, because every bottleneck carries a different weight. The winner rises to the top on evidence.
Run before any build, so you never spend 40 hours of CEO time on a fix a platform already sells.
Turns 30-hour processes into 3-hour processes by redesigning the whole flow from zero with AI at the center.
Layer 3 graduates the skill into a Bottleneck Scout routine that surfaces one scored candidate a week while you sleep.
The Bottleneck Score ranks every candidate bottleneck on three dials from 1 to 10: Money (how directly busting it moves revenue), Drag (how much time and capacity it eats weekly), and Bustability (how much of the fix AI can own today). The total out of 30 turns a vague feeling of being stuck into a ranked lineup with one clear winner.
The Evidence Scan reads your past Claude chats and your notes for signals you leave without noticing: the same task described more than once, 'I keep having to' language, work waiting on one named person, and action items that never close. A bottleneck that appears in four separate conversations is announcing itself. The scan surfaces it with quotes as proof.
Automating a task speeds up one step of an existing process. Busting a bottleneck redesigns the flow so the pile-up never forms, starting from an AI-first rebuild of the whole thing. The skill's One Piece Rule then targets the single point of greatest drag first, which is how a 30-hour process becomes a 3-hour process.