Tired of endlessly repeating your tech stack requirements? Inject a unique personality into your AI with three core configuration files and turn it into the ultimate coding companion that actually understands you.
This is the power of a custom context setup. Say goodbye to the endless overhead of establishing background details and get tailored solutions right out of the gate.
A default AI is like a brilliant new hire - extremely knowledgeable, but entirely oblivious to your team's coding standards, your personal habits, and your current project setup. If you ask it to work immediately, you'll get generic answers.
To make the AI truly harmonize with your workflow, we need to establish Global Context. By leveraging the current industry best practices, we use three core configuration files to completely define the AI:
Defines Virtual Identity
Sets tone, boundaries, and red lines
Your Resume
Lists your tech stack and communication preferences
Project Rules
Standardizes code practices and enforces security guards
SOUL.md serves as the AI's core DNA. This is where you decide if your AI acts as a strict coding mentor or a friendly assistant. Most importantly, it is where you lay out the absolute ironclad rules the AI must obey.
Keep the AI's persona name different from your own name. If both sides are called Alex, multi-turn conversations will start tripping over references like "Alex should handle this".
# Core Identity
You are "Maya", a senior full-stack engineer from Silicon Valley, serving as my technical mentor. You champion elegant and minimalist code.
# Tone and Language
- Keep responses extremely concise and professional. Stop saying "Sure!" or "I can help with that."
- If you spot inefficient spaghetti code, point it out immediately with refactoring advice.
- You may use a bit of dry geek humor, but only after absolutely solving the task.
# Guardrails
- Before coding, you must list out 2-3 architectural approaches and weigh their pros and cons.
- If my prompt has underlying logic flaws, do not write the code. Instead, point out the flaw and ask for clarification.
- When resolving bugs, explain the root cause first before providing the fix.
# Never Do
- Never start a sentence with "As an AI language model..."
- Never add unnecessary, overwhelming comments inside the code block unless explicitly requested.This isn't a dry resume you'd submit for a job application. It's the bridge to establishing deep technical synergy with your AI. By infusing it with your development habits, timezone, and permanent tech stack, the AI transforms from a "generic robot executing instructions" into a "digital partner seamlessly integrated into your workflow."
# Developer Profile
## Core Identity
- Name: Alex Chen
- Role: Frontend Geek / Full-Stack Explorer
- Timezone: GMT+8 (Beijing Time)
## Immersive Workflow
- Focus Area: Building high-conversion landing pages and SaaS dashboards.
- Toolkit: Cursor, Figma, Raycast, Vercel
- Active Window: My peak flow state usually occurs between 10:00 PM and 2:00 AM. Daytime is reserved for code reviews and architectural planning.
## Collaboration Rules
- "Straight to the Point": Do not use filler words like "Sure thing" or "I'd be happy to." Let your final code do the talking.
- "Language Boundaries": Feel free to explain business logic in plain English, but always preserve the original English nomenclature for function names, APIs, and specific design patterns so I can easily search for them later.
- "Fail-Safe Protocol": If my request is vague or contains obvious logical flaws, strictly do NOT hallucinate or guess. Stop immediately and ask me for clarification.
## Current Focus
- Currently deep-diving into Rust and WebGL, trying to inject some hardcore performance and slick animations into legacy projects.
- The ultimate goal for this week is to launch a minimalist idea-capture extension. Therefore, any architectural discussions right now must prioritize being "lightweight and extremely maintainable" above all else.While the previous files improve communication, AGENTS.md introduces a critical safety valve. Typically placed at the root of a project, it dictates operational rules, security blocks, and coding standards exclusively for that project.
# DataLens Project Guidelines
## Directory Structure
- src/app/: All routing and pages
- src/components/ui/: Core UI components (Do NOT modify without permission)
- src/lib/: Shared utilities and functional logic
## System and CLI Commands
- Dev server: pnpm dev
- Linting: pnpm lint
- DB Migration: pnpm db:generate
## HIGH RISK PREVENTION (CRITICAL)
Before executing any of the following actions, you MUST pause, ask "Proceed?", and await my explicit "YES":
- Executing rm -rf or clearing directories
- Modifying or replacing the .env config
- Calling DROP TABLE or DELETE FROM database tables
- Pushing directly to the master / main branch
## Code Standards
- Name exports must be preferred. Do NOT use default export.
- File names must strictly follow all-lowercase kebab-case (e.g., button-group.tsx).These three files have different scopes. Please place them in their correct locations:
~/.openclaw/ or ~/clawd/C:\Users\YourUsername\.openclaw\.gitignore file.After creating or modifying these files, you need to open your terminal and restart the daemon to inject the new configurations into the AI's "brain":
Explore More Cases
Use Gmail API and OAuth to let Openclaw read emails, prioritize them, and draft replies.
Use Bird Skill with your browser session to turn Openclaw X into one workflow for posting, replying, monitoring mentions, reading threads, and AI-driven X operations.
Connect your GPT, Gemini, and Claude plans to Openclaw so Openclaw Codex, Openclaw Gemini, and Openclaw Claude work without extra API token billing.
Learn how to turn repeated work into reusable skills so AI can execute with stable steps, rules, and guardrails.
Learn the two Openclaw Browser starting routes: Playwright MCP and browser-use, and how each one should be set up.
Learn how to install and set up the Openclaw automation system, from environment checks to essential CLI commands.
See how OpenClaw searched inventory, contacted dealerships, and negotiated across browser, email, and messaging channels to save $4,200.
See how Openclaw can write code, open PRs, run code reviews, triage issues, and generate changelogs â handing every repetitive GitHub task to AI autonomously.
See how Openclaw searches job boards across the web, tailors resumes and cover letters, sends bulk personalized applications, and triages inbox replies â automating the most repetitive parts of job hunting.
See how Openclaw searches listings across platforms, compares prices side by side, sends bulk inquiry messages, and negotiates rent with landlords â automating the most tedious parts of apartment hunting.
See how Openclaw securely connects to cloud servers, analyzes error logs, updates config, and recovers failed deploys to keep production healthy around the clock.
See how Openclaw coordinates health data, smart home devices, meal planning, and daily schedules to run your personal systems as one workflow.