The gist
Your Claude toolkit
Foundation session · Wednesday 2 September 2026
Start here
The four part ask
Four things in the ask, and most weak answers are missing one of them.
Task
A comparison table of every screen funding agency in Australia
Audience
A client project team, or your own working file
Standard
Named columns, one row per fund, a source link on every row, and a gap marked as a gap rather than filled in
Boundary
No client material, and nothing you would not put in an email to a stranger
If you are not sure you have given enough, add a line at the end asking it to ask you anything it needs before it answers.
The method
CHART, how to run any task well
C · Chart the task
What you want, who it is for, what good looks like.
H · Hand over clearly
The task, the context, the format, an example.
A · Assess
Is it right? Anything missing? Would you act on it?
R · Refine
Push back. Ask for a change. Try another angle.
T · Track
Know what you put in. Own the output. Keep personal data out.
Plans
Your plan
Prices current at 2 September 2026. Seat rules have changed twice this year, so confirm at signup.
Free
- Price
- $0
- Usage
- Baseline usage
- Claude Code
- No
- Terms
- Consumer terms
- Data processing agreement
- No agreement
Pro
Start here- Price
- USD $20 a month, or $17 annual
- Usage
- 1x
- Claude Code
- Yes
- Terms
- Consumer terms
- Data processing agreement
- No agreement
Max
- Price
- USD $100 for 5x, $200 for 20x
- Usage
- 5x or 20x
- Claude Code
- Yes
- Terms
- Consumer terms
- Data processing agreement
- No agreement
Team
- Price
- USD $25 a seat, or $20 annual, two seat minimum. Premium seat $125, or $100 annual
- Usage
- Standard 1.25x Pro, premium about 6.25x
- Claude Code
- Yes on every seat
- Terms
- Commercial terms
- Data processing agreement
- Data processing agreement included
The contract and the capacity pull against each other. Pro and Max give you capacity on consumer terms with no agreement behind them. A Team standard seat gives you the agreement but only 1.25 times Pro, which is thin for sustained Claude Code work. Both at once means a Team premium seat, and the two seat minimum.
Privacy
The training setting
Anthropic's help centre and its Privacy Policy describe the default differently. Open Settings, then Privacy, and read what your own account says.
- The control is called Model Improvement.
- Left off, your conversations are held for 30 days and are not used to train the models.
- Turned on, they are kept for up to five years and are used.
- One carve out either way: a conversation flagged for safety review can be used.
- Incognito chats are never used.
- Claude for Work and the API sit under separate terms.
Sources
The policies themselves
The Privacy Center. Consumer articles for Free, Pro and Max.
The Privacy Policy. Effective 8 July 2026.
The Consumer Terms. Effective 8 October 2025.
The Data Processing Addendum. Attaches to the commercial terms only.
"This Data Processing Addendum ("DPA") is incorporated into and forms part of the Anthropic Commercial Terms of Service or other agreement between Customer and Anthropic that references this DPA and governs Customer's use of the Services"
That sentence is the whole point. The agreement attaches to the commercial terms, so a consumer plan does not have one.
Boundaries
What stays out
Your own methods and IP
What protects them is the setting above, the plan you are on, and the fact that a model returns the middle. Your method is not the middle. What you put in does not come back out to someone else through the model.
Your clients' material
A different question, and a decision each time. What is in the engagement. What the client would say if you asked them. Whether the same work runs on a de-identified version. Signed contracts and any legal matter on foot stay out, along with anyone's personal records.
And the one no setting solves: do not ask it to generate First Nations cultural material. That is not a privacy problem, it is a different problem.
Models
Which model to use
Think of it as bike gears. You stay in the middle gear almost all of the time.
Sonnet
Everyday, fast and strong. Stay here.
Opus
Hard, messy problems where you want it to think longer.
Haiku
Quick, light jobs like tidying a list or renaming things.
Setup
Set Claude up around you
The two folders on your desktop
My Claude DNA holds work-dna.md and ways-of-working.md, so your profiles are yours to reuse in any project you ever build. My Work Tasks holds the templates, examples and previous work you pull from. Both live on your own computer, not inside Claude.
Your project
One room for the work you do often. Instructions hold the short signpost, Files hold your profiles and your reference material.
Your two profiles
work-dna.md, what you do and how the work runs. ways-of-working.md, how you think and how you want things handed back. Save both as markdown files on your own computer, then load them into the project.
Copy and run
The prompts
The Work DNA interview
Paste into a new chat. Claude interviews you, one question at a time, for about fifteen minutes. It writes the file at the end. Save it as work-dna.md.
You are going to interview me to build my Work DNA. This is a short profile of who I am at work, how I like things written and worked, and the tasks I want help with. I will save it into my Claude project so you are primed with me from my first message. How to run this interview: - Ask me about ONE area at a time, in the order below. Do not send me all the questions at once. - After each answer, ask me one or two follow-ups to get the specifics, then move to the next area. - Keep it plain and friendly, like a colleague getting to know my job. If you use a term I might not know, explain it in passing. - Do not ask me for signed contracts, confidential documents, or anyone's personal records. If I start to share those, remind me to keep them out and to keep it general. - When we have been through every area, build my Work DNA as a downloadable markdown file, using the headings listed at the end, ready for me to save and load into my project. Keep my own words where they matter. Name the file work-dna. Interview me across these areas, in this order. 1. Who I am at work - What is your name, and what would you like me to call you? - What is your role, and what is the business? - Who are your clients, and what do they come to you for? - How long have you worked this way, and what did you do before it? 2. What I actually do - In your own words, what is your work mainly about, and what are you responsible for? - Walk me through a normal week. What comes up every day, every week, and every month? - Which tasks eat the most of your time? - Which tasks do you find a slog, or tend to put off? - What are you the go-to person for? - Is there a cycle that changes your work, funding rounds, reporting periods, board meetings, festivals? 3. The people and systems I work with - Which people or organisations do you deal with, and what do you go to each of them for? - What systems and tools do you use day to day? - Where does your information usually live, a particular folder, an inbox, a drive? 4. What I produce, and what good looks like - What do you actually make or send, strategies, reports, submissions, briefings, facilitation plans, summaries? - For each main one, who reads it, and what does a good one look like, how long, what format, what tone? - Are there formats or templates you have to follow? - If you have one you are happy with, describe it, or paste a de-identified example. 5. How I like things written, my voice - When something is written for you, how do you like it, plain and short, or fuller? Relaxed or formal? - Bullet points, or flowing sentences? - Any words or phrasings you cannot stand, or always want? - If you have a short piece of your own writing, paste it so I can match how you write. Take any names out first. 6. How I like to work with you - When I give you a job, do you want me to ask questions first, or have a go and show you a draft? - Do you want a recommendation, or a few options to choose from? - Do you like a short answer, or the full working? - Step by step, or just the result? 7. What I want to make faster - If I could take two or three jobs off your plate, what would they be? - Which of those do you do most often? - Which one, made faster, would give you back the most time or the most peace? When the interview is done, build my Work DNA as a downloadable markdown file named work-dna, under these headings: - Role and context - What I do and the rhythm of my work - The people and systems I work with - What I produce, and what good looks like - How I like things written (my voice) - How I like to work with Claude - What I want to make faster (my priorities) Begin with area one. Ask me the first question now.
The Ways of Working interview
Straight after the first one, in a fresh chat. About ten minutes. It stays at work and does not ask about your life. Save what it writes as ways-of-working.md.
You are going to interview me about how I like to work, and how I want you to work with me. This is my Ways of Working profile. It sits alongside my Work DNA and helps you match how I think and communicate, not just what my job is. I will save it into my Claude project. Stay at work. Do not ask me about my family, health, money, or personal history. If something can only be answered by going into my private life, skip it. This is about my working style only. How to run this interview: - Ask me about ONE area at a time, in the order below. Do not send me all the questions at once. - After each answer, ask one light follow-up if it helps, then move on. - Keep it plain and easy, like a colleague working out how to work well with me. - When we are done, build my Ways of Working profile as a downloadable markdown file, under the headings at the end, ready to save and load into my project. Keep my own words where they matter. Name the file ways-of-working. Interview me across these areas, in this order. 1. How I think and take things in - When you are getting your head around something, do you want the big picture first, or the detail first? - Do you like to work one thing at a time, or hold a few threads at once? - Does a wall of text help you or lose you? 2. How I like to be communicated with - Do you prefer it direct and to the point, or warmed up a little? - When there is a decision, do you want a clear recommendation, or a few options to weigh up? - Short answers, or the full reasoning? 3. What drains me and what lifts me at work - Which kinds of tasks drain you, the ones you would gladly hand off? - Which kinds of work do you find satisfying, the ones you want to keep doing yourself? - This tells me what to take off your plate and what to leave with you. 4. How I handle pressure and deadlines - When things are busy, do you want me to keep it tight and fast, or steady and thorough? - Do you like a job broken into steps, or handed back finished? - Is there a time of day or a kind of week when you want me to keep things simple? 5. What I value in good work - When something comes back to you, what makes it good in your eyes, accuracy, speed, polish, plain language, something else? - What is an instant no for you, what makes you think this is not right? 6. How I want you to show up for me - In a sentence or two, what do you most want me to be for you, a sounding board, a drafter, a checker, a planner, a second set of eyes? - Is there anything you want me to always do, or never do? When the interview is done, build my Ways of Working profile as a downloadable markdown file named ways-of-working, under these headings: - How I think and take in information - How I like to be communicated with - What to take off my plate, and what to leave with me - How to pace things under pressure - What I value in good work - How I want Claude to show up Begin with area one. Ask me the first question now.
Generate your own prompts
Run this inside your project, once both profiles are loaded. It writes the prompts you will actually reuse.
You are a prompt coach. Below is my Work DNA. Read it closely. First, privately work out my role, my main tasks, the tools and people I deal with, and the two or three jobs I most want to make faster. Do not show me this working. Then write me ten ready-to-use prompts I can paste straight into Claude for my actual work. Rules for the prompts: - Plain English, written the way I would actually type, no jargon. - Each one has a short, clear title, and a one-line note on when to use it. - Put anything I would swap out in [square brackets], no more than four per prompt. - Tie them to my real tasks, my real tools, and the jobs I want made faster. - Mix everyday quick ones with two or three bigger ones. - Keep my protected material out: no contracts, confidential documents, or personal records. Build me the ten prompts as a downloadable markdown file named my-prompts, so I can save it and load it into my project. After the file, tell me which two or three you would turn into a saved Skill first, and why. [Paste your Work DNA here if it is not already in this project.]
The funding scan, rebuilt
The version we ran on the day. Keep it, because the next time an agency changes a fund you can run it again in ten minutes instead of a week.
I need a complete comparison of what every Australian screen funding agency offers for online and digital content creators. Task. Build one table, one row per fund. Columns: agency, fund name, amount available, who can apply, rounds per year, eligibility summary, source URL. Scope. Every federal, state and territory screen agency in Australia. Include Screen Australia and every state and territory equivalent. If an agency has no fund that fits, give it a row and mark it as none, do not leave it out. Standard. Every row carries a source URL to the page the figure came from. Where a figure is not published, write not published rather than estimating. Where a fund is open to online and digital content but not exclusively, say so in the eligibility column rather than excluding it. Boundary. Publicly available information only. Do not infer amounts from past rounds and present them as current. Use Research for this. When you are done, tell me which rows you are least confident about and why.
Off the platform
What comes next
Everything in the session happened inside Claude. The next step goes the other direction: Claude reading the folders on your own machine.
A workspace you own
A folder on your computer, one folder per client, one per job. It would still make sense if you never used Claude again.
A file that describes how you work
A CLAUDE.md at the top of that workspace, read every single time, so the standing instructions are not retyped. Your two profiles go into it nearly as they are.
Claude Code, which is a builder rather than a robot
It reads and writes files on your own computer while you watch. It does nothing unless it is open and you have asked. The folder you open is the boundary, and that is the whole security model.
Markdown, and why
A plain text file with light formatting. A hash makes a heading, a dash makes a bullet. It opens anywhere, it will still open in ten years, and it is exactly what your two profiles already are.
Practice
Your first week
Day 1
Open your project and do one real piece of client work inside it.
Day 2
Run your two profiles over a second piece of work and fix what they get wrong.
Day 3
Put one live task through the four part ask before you start it.
Day 4
Ask it for the thing you would normally drop at the detail stage.
Day 5
Start the AI Fluency course, no charge.
Next
Keep learning, no charge
Anthropic's own course, and it carries on from where the session finished.
https://anthropic.skilljar.com/ai-fluency-framework-foundations