AI Guides › Step-by-step guides

Five Claude Setups That Remove Repeated Chores: Instructions, Project, Skill, Connector, Artifact

By Nigel Guy · 6 min read

Most people open a fresh Claude chat, re-explain who they are and what they want, get a decent answer, and repeat tomorrow. Nothing is wrong with the chat. The mistake is treating every task as new when Claude has five places to store the parts that do not change. This guide is written as a standalone "part two": we could not see the original post it was commissioned from, so we have not assumed what its part one covered, and we have built everything below from Anthropic's help pages as checked on 2026-10-04.

The rule: set up a thing only when you have explained the same thing three times. Match the repeated annoyance to the right container, and do one this week, not all five.

Before you start

The Annoyance Table

The annoyance The container Plan note Step
"I always have to say how I like answers" Account-wide instructions All plans 1
"I keep pasting the same background files" A Project All plans; Free limited to five projects 2
"I keep re-specifying the same task" A custom Skill Free, Pro, Max, Team, Enterprise 3
"I keep copying data in from another app" A connector Free allows one custom connector 4
"I keep rebuilding the same little tool" An artifact All plans create them; storing data and connecting apps are paid 5

Step 1 — Write your standing instructions

Anthropic's help page places account-wide instructions under your initials (lower left) > Settings. These apply to every conversation, so they suit only what is true everywhere: spelling (British, if that is you), how blunt you want it, the sort of work you do, and terms you use.

  1. Open Settings and find the personal instructions field.
  2. Write five to eight short lines. Say what you do, how you like answers laid out, and what to avoid.
  3. Save, then start a new chat and ask something unrelated to see them applied.

While you are in Settings, open Memory. Per the help page, memory is on by default for Free, Pro and Max, and you can read, edit or delete saved topics there or tell Claude "remember this" or "forget that". Incognito chats (ghost icon, top right) are not saved to history or memory.

Step 2 — Put the background in a Project

A Project is a workspace with its own instructions, files and chat history. Free accounts can have up to five. Paid plans get a retrieval mode that Anthropic says expands capacity by up to 10x. Sharing a project is Team and Enterprise only.

  1. Create a project from the Projects area and name it after the job, not the client.
  2. Add project instructions: role, audience, format, what to refuse.
  3. Upload two or three files that define good: a style guide, a price list, a past piece you were proud of.

Use this prompt to draft the instructions, then paste the result in.

Role: you are an expert at writing standing instructions for an AI assistant.

Context: I will use one Claude Project for [JOB_DESCRIPTION]. The audience is [AUDIENCE]. Good output looks like [WHAT_GOOD_LOOKS_LIKE]. Things I never want: [DEAL_BREAKERS].

Task: ask me up to five questions if anything above is missing or vague. Do not guess. Then write project instructions in this order: role, audience, output format, rules, what to refuse or flag.

Constraints: British English, under 250 words, no filler, no invented facts about me or my business.

Before you answer, check that every rule is something you could actually follow and that none contradict each other. Output only the instructions.

Fill in the four bracketed items.

Step 3 — Turn a repeated task into a Skill

Skills are for tasks with a fixed method: the same steps and the same output shape every time. Anthropic's help page gives the route: Customize > Skills, click "+", then "Create skill" > "Upload a skill", and upload a ZIP of a skill folder. Built-in skills for Excel, Word, PowerPoint and PDF are used automatically when relevant.

  1. Pick one task you did at least three times this month.
  2. Write the steps once, with an example of a good result, in a skill folder following Anthropic's structure.
  3. ZIP it, upload it, then test with a real task.

Only install skills from sources you trust, and read the files first, especially anything that runs code.

Step 4 — Connect one app, read-only first

Connectors let Claude read from other apps instead of you pasting. On Pro and Max the path is Customize > Connectors > "+ Add" (custom ones: "Add custom connector"). In a chat, use the "+" button lower left > Connectors to toggle them on or off. Free accounts are limited to one custom connector.

  1. Choose the one app you paste from most.
  2. Add it and sign in. Prefer "Sign in when needed" if offered.
  3. Ask for a read-only task first, such as summarising this week's items.

Anthropic warns that malicious servers can embed hidden instructions. Connect only to servers you trust and read the permissions requested.

Step 5 — Keep a small tool as an artifact

An artifact is anything Claude makes that you would put in front of someone: a document, a dashboard, a small interactive tool. On paid plans artifacts can store data (20 MB limit per the help page) and connect to apps. Note that Anthropic flags changes to publishing and sharing for artifacts created after 16 September 2026, so check the current sharing options before promising anyone a link.

  1. Describe the tool: inputs, what it calculates, what it shows.
  2. Try it with real numbers, ask for changes in chat, and export if you need a file.
  3. Keep it private until you are sure what it exposes.

Check it worked

For each step, run one cold test in a new chat.

Step Test Pass
1 Ask an unrelated question Tone and spelling match
2 Ask a job question with no background It uses your files without prompting
3 Run the task with a new input Same structure as your example
4 Ask a read-only question Real data, correctly cited
5 Reopen next day Tool and data are there

What to skip

Guardrails

Sources

All 751 AI guides · JulieMango plans from £17/mo