Claude Custom Instructions: Examples You Can Adapt
Summary
- Claude custom instructions are reusable preferences you set once so Claude follows your style, format, and constraints across chats.
- This guide gives adaptable instruction templates for writing, coding, research-style synthesis, and business workflows.
- You will learn how to give Claude instructions that are specific, testable, and resilient when tasks change.
- Each example includes optional “tighten/loosen” variants so you can control verbosity, tone, and risk.
- You will also see a practical way to save, find, and reuse your best instruction sets with CopyCharm when you work across tools.
Adaptable Claude custom instruction examples for tone, structure, and constraints, plus quick steps to write instructions Claude can follow consistently.
What “Claude custom instructions” are (and what they are not)
Claude custom instructions are persistent guidance you provide so Claude knows how you want it to respond: your preferred tone, formatting, depth, and boundaries (for example, “ask clarifying questions first” or “use bullet points and include a checklist”). The goal is consistency across many conversations, so you do not have to restate the same preferences every time.
They are not a substitute for task-specific prompts. Even with strong custom instructions, you will still get better results when you add context in the moment: audience, constraints, examples, and what “done” looks like.
Version/account differences: Where and how you set custom instructions can vary by Claude plan, app (web vs desktop), and updates. If you do not see a dedicated “custom instructions” area, look for settings related to “preferences,” “profile,” “response style,” or similar. If your Claude experience does not support persistent instructions, you can still use the same content by pasting it at the top of a new chat as a “system-style” preface.
How to give Claude instructions that actually stick
If you want Claude to follow instructions reliably, write them like a compact spec. These patterns help:
- Make preferences observable: “Use headings and bullets” is easier to follow than “be organized.”
- Separate defaults from exceptions: “Default to concise; if I say ‘deep dive,’ expand.”
- Define the interaction loop: “Ask up to 3 clarifying questions when requirements are ambiguous.”
- Set boundaries: “If you are unsure, say what is unknown and offer options to verify.”
- Specify output formats: “Return JSON with keys X/Y/Z” or “Provide a table plus a short recommendation.”
A reusable custom-instructions template (copy/paste)
Use this as your base, then swap the bracketed parts.
Claude Custom Instructions (Template)
- Role: You are my [role: editor / analyst / tutor / pair programmer].
- Audience: Write for [audience] with [reading level / domain knowledge].
- Default style: [concise/neutral/friendly], avoid [jargon/marketing tone], use [bullets/headings].
- Process: If the request is ambiguous, ask up to [1-3] clarifying questions before answering.
- Quality bar: State assumptions, list tradeoffs, and include a quick example when helpful.
- Boundaries: If you are uncertain, say what is unknown and provide safe next steps instead of guessing.
- Output rules: When I ask for [emails/specs/plans], format as [structure].
Claude custom instructions: examples you can adapt
Each example below is written as a “set-and-forget” instruction block. You can paste one into Claude’s custom instructions (or into the first message of a new chat if you are using a workaround).
1) Clear, structured writing (general-purpose)
Instruction block
- Write in clear US English for a general audience.
- Start with a 1-2 sentence direct answer, then add details.
- Use headings and bullet points; keep paragraphs under 5 lines.
- When giving steps, include prerequisites and a quick “check your work” section.
- If I provide examples, mirror their tone and level of detail.
Make it tighter: “Default to under 250 words unless I ask for ‘long-form.’”
Make it deeper: “Include edge cases and a short example for each major point.”
2) Executive summaries for busy stakeholders
Instruction block
- Assume the reader is a busy executive.
- Lead with: Decision, Why, Risks, Next steps (in that order).
- Use short bullets; avoid long background unless asked.
- Call out unknowns explicitly and list what would change the recommendation.
Optional add-on: “End with a 1-line ‘If you only remember one thing’ takeaway.”
3) Marketing copy with guardrails (no hype)
Instruction block
- Write benefit-led copy without exaggerated claims.
- Avoid superlatives and guarantees; use qualified language (can help, may, designed to).
- Prefer concrete specifics: who it is for, what it does, what it does not do.
- Offer 3 variants: conservative, neutral, and bold (still truthful).
Optional add-on: “If you mention numbers, ask me to confirm them first.”
4) Technical explanations for mixed audiences
Instruction block
- Explain technical topics for a mixed audience (some technical, some not).
- Start with a plain-English overview, then a deeper technical section.
- Define acronyms on first use.
- Use a small example (pseudo-code or a concrete scenario) when it clarifies the concept.
5) Pair-programming and code review
Instruction block
- Act as a careful pair programmer.
- Before coding, confirm: language/runtime, constraints, and expected inputs/outputs.
- Prefer small, testable functions and explain key decisions briefly.
- When reviewing code, list: correctness issues, edge cases, readability, and performance considerations.
- If you are unsure about an API detail, say so and suggest how to verify.
Optional add-on: “When you propose changes, show a minimal diff-style snippet.”
6) Research-style synthesis (without pretending to browse)
Instruction block
- When I ask for research or comparisons, separate: known facts (from what I provide) vs assumptions.
- Ask for missing inputs that materially affect the answer.
- Provide a balanced view with tradeoffs and decision criteria.
- Do not claim you accessed sources unless I pasted them into the chat.
7) Brainstorming that stays practical
Instruction block
- Generate ideas that are feasible for a small team and limited budget.
- For each idea, include: who it is for, why it works, first step, and a likely risk.
- After listing ideas, recommend the best 2-3 based on my constraints.
8) Meeting notes to action items
Instruction block
- Turn my raw meeting notes into: summary, decisions, open questions, and action items.
- Action items must include an owner placeholder, due date placeholder, and clear next step.
- If something is ambiguous, list it under “Open questions” rather than guessing.
9) Sales emails that respect compliance and reality
Instruction block
- Write concise sales emails with a respectful tone.
- Do not invent customer names, metrics, or case studies.
- Include one clear CTA and one low-friction alternative (e.g., “Should I send details?”).
- Provide a subject line and a 2-sentence follow-up.
10) “Ask first” mode (reduces wrong assumptions)
Instruction block
- If my request could mean multiple things, ask clarifying questions before answering.
- Ask no more than 3 questions at a time.
- If I say “no questions,” proceed with explicit assumptions and label them.
Practical ways to test and refine your instructions
Custom instructions work best when you iterate. Here are quick tests you can run without overthinking it:
- Consistency test: Ask Claude the same question in two separate chats. If the structure changes, tighten formatting rules (headings, bullets, order).
- Constraint test: Give a task with a hard limit (word count, JSON schema, “no jargon”). If Claude drifts, restate the constraint as an output rule.
- Ambiguity test: Give an underspecified request. If Claude guesses too much, add an “ask-first” clause.
- Over-verbosity test: If answers are too long, add a default length and a trigger phrase for longer output (“deep dive”).
Limitations to expect (so you are not surprised)
- Instructions can conflict: If you ask for “extremely concise” and also “include examples and edge cases,” Claude may compromise. Decide what matters most and state priorities.
- Task context still matters: Custom instructions do not replace providing inputs (your draft, your data, your constraints).
- Platform behavior can change: Claude’s UI and instruction handling can vary by plan/app and may evolve. Keep a copy of your best instruction blocks so you can reapply them if needed.
Saving and reusing your best Claude instruction sets with CopyCharm
If you iterate on custom instructions, you will likely end up with multiple “instruction sets” (for example: one for executive summaries, one for code review, one for marketing copy). CopyCharm can help you keep those blocks easy to reuse across chats and tools.
A concrete workflow (save, find, reuse)
- Save: When you finalize an instruction block you like, copy it and save it in CopyCharm as a Saved Prompt (separate from favoriting a copied clip).
- Find: Later, search your saved prompts in CopyCharm using a memorable phrase you included inside the block (for example, “Decision Why Risks Next steps” or “ask up to 3 clarifying questions”).
- Reuse: Copy the saved prompt back into Claude when you update your custom instructions, start a new project, or want a different “mode” for a specific chat.
This is also useful when you work across multiple apps: for Claude (and other tools like Gemini, Cursor, email, or documents), the verified workflow is manual reuse: search/retrieve in CopyCharm, then copy/paste into the destination application.
When ChatGPT access matters (and what it can and cannot see)
If you also use ChatGPT, CopyCharm includes an authenticated ChatGPT connector backed by optional AI Access sync. After you sign in with an eligible active CopyCharm purchase, authorize the CopyCharm Desktop connection, enable and complete AI Access sync, and authorize the ChatGPT connector, ChatGPT can search or list recent supported synced items and retrieve a selected item’s full text. ChatGPT can only access supported Synced Data (Favorite Clips, Saved Prompts, and optional Other Clips within your selected time range); it cannot access unsynced local CopyCharm data.
Try CopyCharm for saving and reusing instruction blocks
Frequently Asked Questions
FAQ 1: What are Claude custom instructions used for?
Answer: They are used to set persistent preferences like tone, formatting, depth, and boundaries so Claude responds in a consistent way across conversations. You can treat them as your default “house style” and working rules.
Takeaway: Use custom instructions for defaults; use prompts for the specific job.
FAQ 2: How do I give Claude instructions so it follows them consistently?
Answer: Make instructions specific and testable: define the output structure (headings/bullets), set a default length, add an “ask clarifying questions” rule, and state what to do when uncertain (label assumptions instead of guessing). Then run a quick consistency test in a new chat and refine the wording.
Takeaway: Treat instructions like a mini spec, not a vibe.
FAQ 3: Should I put task details in custom instructions or in the prompt?
Answer: Put stable preferences in custom instructions (tone, formatting, how to handle ambiguity). Put changing details in the prompt (audience for this piece, your constraints, your source text, what you want delivered today). If you mix them, prioritize clarity: “Defaults are X; for this task do Y.”
Takeaway: Defaults live in instructions; specifics live in the prompt.
FAQ 4: What is a good Claude instructions example for concise answers?
Answer: Use a default length rule plus an expansion trigger, such as: “Start with a 1-2 sentence direct answer. Keep total output under 200 words unless I say ‘deep dive.’ Use bullets for steps.” This keeps brevity without blocking detail when you need it.
Takeaway: Combine a short default with a clear “go longer” keyword.
FAQ 5: What is a good Claude instructions example for coding help?
Answer: Add an “ask-first” checklist and review criteria: “Confirm language/runtime and expected I/O before coding. Prefer small functions. When reviewing, list correctness issues, edge cases, readability, and performance considerations. If unsure about an API detail, say so and suggest how to verify.”
Takeaway: Good coding instructions define the workflow, not just the output.
FAQ 6: How do I write instructions that reduce hallucinations or guessing?
Answer: Add explicit uncertainty handling: “If you are not sure, say what is unknown, list assumptions, and offer safe next steps to verify.” Also tell Claude not to invent specifics like metrics, quotes, or sources unless you provide them in the chat.
Takeaway: Tell Claude what to do when it does not know.
FAQ 7: Can I maintain multiple instruction sets for different workflows?
Answer: Yes. Many people keep separate instruction blocks for different modes (executive summaries, brainstorming, code review, marketing copy). If Claude only supports one persistent set in your current interface, you can keep alternates as copy/paste blocks and apply them at the start of a chat when needed.
Takeaway: Keep “modes” as separate blocks so you can swap quickly.
FAQ 8: Can CopyCharm help me reuse Claude instruction blocks across chats?
Answer: Yes. CopyCharm is a Windows desktop app that saves copied text locally and lets you search past clips, favorite important clips, and separately save reusable prompts. You can store your best Claude instruction blocks as Saved Prompts, search them later, and copy/paste them into Claude. For Claude specifically, reuse is manual copy/paste; CopyCharm’s authenticated connector is for ChatGPT and can only search/retrieve supported Synced Data after eligible authorization and AI Access sync.
Takeaway: Save instruction sets once, then retrieve and paste them when you switch tasks or chats.