Claude Tips & Tricks: The Settings, Projects, and Skills That Make Claude Actually Useful
Claude is one of the most capable AI assistants available, but most business users only use 10% of what it can do. This guide covers the org-level settings, project setups, custom skills, and pre-built agents that turn Claude from a chat tool into a serious work partner.
Why Most People Get Claude Wrong
Most Claude users treat it like a smarter version of Google. They open a chat, ask a question, get an answer, close the tab.
That works, but it leaves the most valuable Claude features completely unused — the same features that turn Claude into a personalized assistant that already knows your company, your role, your writing style, and the way you actually work.
This Claude AI guide walks through the four configuration layers that experienced Claude users rely on every day: organizational instructions, project instructions, custom skills, and pre-built agents.
Setting these up takes 20 to 30 minutes once. After that, every conversation gets better automatically.
Setting Up Your Org Instructions in Claude
Org-level instructions tell Claude how to behave across every conversation you have. This is the highest-leverage 10 minutes you can spend with the tool.
Where to Find Org Instructions
In Claude, navigate to Settings → Profile → Personal Preferences (sometimes called "What should Claude call you" and "What should Claude know about you"). These are global settings that apply to every chat unless overridden by a Project or system prompt.
What to Include in Your Org Instructions
Your name, your role, your company, and what your company does. Example: "I'm the Director of Marketing at a 75-person B2B SaaS firm focused on cybersecurity. We sell to mid-market CISOs and CIOs."
Tone, length, format preferences. Example: "Respond in concise, direct prose. No corporate buzzwords. No fluff intros. Use bullets only when the content is genuinely a list, not for visual padding."
Your common work tasks. Example: "I frequently work on positioning, sales enablement copy, and content strategy. Assume I'm a strong writer who needs Claude as a thinking partner, not a writer."
Example: "Push back on weak reasoning. If I propose something flawed, say so. Don't be agreeable for the sake of it."
A Sample Org Instructions Template
I am [Name], [Role] at [Company]. We are a [size] [industry] company that does [one-line description]. My typical work involves [list 3-4 common tasks]. When responding to me: - Be direct and concise. No filler. - Avoid corporate jargon and AI-sounding phrases. - Use bullets only for genuine lists. - Push back on weak reasoning or flawed assumptions. - Default to plain English over technical vocabulary unless I ask for depth. Things I care about: - [Personal preference 1] - [Personal preference 2] - [Personal preference 3]
Copy this, customize it, paste it into Claude's personal preferences. Done.
Setting Up Project Instructions in Claude
Claude Projects are the second layer of personalization. While org instructions apply to every chat, Project instructions apply only to conversations inside a specific Project.
What a Project Is
A Project in Claude is a workspace that contains:
- Custom instructions specific to that work
- Uploaded files Claude can reference (the "Project Knowledge")
- All conversations related to that work, grouped together
Think of a Project as Claude's specialized memory for a specific area of your job.
When to Create a Project
Create a Project for any recurring area of work where you want Claude to have consistent context. Common examples:
- One Project per major client (if you do client work)
- One Project per ongoing initiative (a product launch, a hiring plan, a fundraising round)
- One Project per recurring deliverable (weekly board update, monthly newsletter, sales follow-ups)
- One Project per role-based workflow (sales prospecting, content production, financial analysis)
What to Include in Project Instructions
Project instructions should cover what's specific to that work, not what's general to you. Examples:
What to Add to Project Knowledge
The Project Knowledge area lets you upload documents Claude can reference. Upload your most-used reference material:
- Brand voice or style guides
- Positioning documents
- Past examples of work you want Claude to match
- Internal SOPs or playbooks
- Customer personas or ICP documents
- Past meeting transcripts or strategy docs
Claude will pull from these automatically when relevant.
Creating a Skill in Claude
Skills in Claude are reusable mini-instructions you can invoke inside any conversation. Think of them as named templates for repeated tasks.
What a Skill Looks Like
A skill is essentially a saved prompt with a name. Instead of writing out the same 200-word prompt every time you need a meeting summary, you save it as a Skill called "Summarize Meeting" and invoke it whenever you need it.
Five Skills Worth Creating Today
If you do any of the following weekly, build a Skill for it:
Meeting Summary Skill
A reusable prompt that takes a transcript and produces a structured summary with decisions, action items, owners, and unresolved questions.
Status Update Skill
A reusable prompt that takes raw notes from your week and produces a clean status update in your team's preferred format.
Email Reply Skill
A reusable prompt that takes an incoming email and drafts a reply in your voice with three different tone options (formal, neutral, warm).
Strategic Memo Skill
A reusable prompt that takes 5 bullet points and turns them into a board-ready strategic memo.
Critique Skill
A reusable prompt that takes any piece of writing or strategy and gives sharp, structured criticism — what's weak, what's missing, what's redundant.
How to Build a Skill
In Claude, navigate to the Skills area (location varies by interface). Click "Create Skill." Give it a clear name (e.g., "Meeting Summary"). Paste the full prompt you want to run. Save. From then on, invoking the Skill in any chat applies that prompt's instructions to whatever input you provide.
Loading Pre-Built Agents in Claude
Beyond basic chat, Claude supports more advanced agentic workflows where Claude can take actions, use tools, and operate over multiple steps.
What an Agent Is in Practical Terms
An AI agent extends Claude with the ability to take actions and pursue goals across multiple steps, rather than just respond to a single message. For example, a research agent might:
- Take your topic
- Search the web for information
- Read multiple sources
- Compare findings
- Produce a structured report
A pre-built agent is a saved configuration that handles a specific multi-step workflow you don't want to set up from scratch each time.
Common Pre-Built Agents Worth Exploring
Depending on your Claude tier and integration setup, common pre-built agents include:
- Research agents — for topic research, competitive analysis, market scans
- Writing agents — for long-form content drafting with iteration built in
- Code agents — for writing, reviewing, and debugging code
- Data agents — for analyzing spreadsheets, CSVs, or uploaded datasets
- Document agents — for processing long contracts, reports, or PDFs
Security Considerations Before Loading Agents
Agents are more powerful than basic chat, which means they introduce more security considerations. Before loading or running an agent that has tool access (web search, file system, API calls), consider:
- What data will the agent see?
- What actions can it take on your behalf?
- Is the agent operating in a sandboxed environment, or does it have access to live systems?
- Who has visibility into what the agent does?
For sensitive workflows, stick to agents from trusted sources (Anthropic-built or vetted partners). For broader use, set up agents in environments with appropriate access controls. Read Using AI Safely at Work for the full guidance.
Common Claude Setup Mistakes
"Be helpful and respond clearly" is useless. Be specific about your role, your work, and how you want Claude to behave.
If something only applies to one client or one initiative, it belongs in a Project, not your global settings. Keeping global settings lean makes them more effective.
The Project Knowledge area is the highest-leverage feature in Projects. If you have positioning docs, style guides, or past examples, upload them. Claude's outputs improve dramatically.
Your first version of org instructions and Project setups will not be the best version. Refine them every few weeks based on what's working and what isn't.
Free-tier Claude has different data handling than Claude Team or Enterprise. If you're working with customer data, contracts, or internal financials, use an enterprise tier.
See Using AI Safely at Work for full data-handling guidance.
Where to Go From Here
Master the prompt fundamentals: Prompt Engineering Guide
Apply Claude to your role: Browse The AI Prompt Library for role-specific Claude prompts
Compare tools: Copilot Tips & Tricks if you also work in Microsoft 365
Use Claude safely: Using AI Safely at Work
Want Your Team Trained on Claude?
Defensible runs custom AI workshops that teach your team how to set up Claude properly for your business, build Projects that actually deliver value, and use Claude safely with sensitive data. From CEO to intern, everyone leaves with a working setup.