This repository contains the system prompt specification that governs AI behavior when working with Lucky, a highly experienced technical professional with deep expertise in systems administration, security, networking, and software engineering.
Current production system prompt: System Prompt v2/Current System Prompt v2.5/lucky-system-prompt-v2.7.md
Use this prompt with:
- Claude Desktop (Windows 11) - primary MCP development environment
- Claude Code (WSL Ubuntu-24.04) - primary coding environment
- OpenCode (Linux servers) - production infrastructure
The AI operates as a peer collaborator, not an assistant. Communication is direct, technical, and efficient. The AI never patronizes, over-explains fundamentals, or hedges with disclaimers about its nature or limitations.
Key principles:
- Speed and pragmatism over perfection
- Code speaks louder than explanations
- When in doubt, implement and iterate
- Working solutions, not academic exercises
- Trust Lucky's experience—he knows what he's doing
Evolution of the v2.x series incorporating lessons from real-world projects.
- lucky-system-prompt-v2.7.md - Current production version
- lucky-chat-gpt-bootstrap-system-prompt.md - Bootstrap prompt for prompt length-limited AIs
Version 2.5 improvements:
- GitHub CLI preference over plain git for GitHub operations
- Mandatory backup/verify/diff workflow for all config files
- Universal tilde ban + absolute paths only
- Per-command user/path verification
- Filesystem context checks before writes
- WSL access procedure for Windows 11
- Directory traversal scope limits (end of prompt)
- Line count: 1393 lines (up from 1173 in v2.4 = +220 lines = +18.7%)
The system prompt is organized into 54 numbered sections covering:
Core Identity and Communication Protocols (1-5): User identity and expertise assumptions, response format/length, tone, question handling, and formatting constraints.
Code and Documentation Standards (6-13): File headers, GitHub CLI preference, critical file edit protocol, path/user verification, filesystem checks before writes, code style, documentation format, and directory entry protocol.
Execution and Permission Boundaries (14-16): Autonomous execution scope, mandatory permission requests, and root-cause-over-hacks philosophy.
Quality Assessment Framework (17-21): Package quality thresholds, discovery methods by tool class, dependency health, native module risk, and path/environment debugging.
Verification Requirements (22-23): File operation verification and API/technical-claim verification requirements.
Platform and Environment (24-37): Operating-system scope, filesystem context rules, cross-filesystem methods, MCP server guidance, WSL access from Windows 11, and Cursor-specific behaviors.
State Preservation and Long Sessions (38): Multi-hour debugging expectations.
Failure Mode Guardrails (40-43): Frustration handling, ambiguity handling, pattern recognition, and build-vs-buy framework.
Technical Decision Framework (44-45): Language/tool selection and architecture preferences.
Anti-Patterns and Explicit Prohibitions (46-48): Communication, execution, and explanation anti-patterns.
Completion Signals (49): Task completion/error/option signaling format.
Environment Conflict Resolution (50-53): Python version isolation, environment verification, process criticality rules, and core operating principles.
Directory Traversal Scope Limits (54): Instructional-file discovery boundaries and cross-tree traversal restrictions.
v2.5 - February 2026
- GitHub CLI preference over plain git for GitHub operations
- Mandatory backup/verify/diff workflow for all config files
- Universal tilde ban + absolute paths only
- Per-command user/path verification
- Filesystem context checks before writes
- WSL access procedure for Windows 11
- Directory traversal scope limits
- Line count: 1393 lines (up from 1173 in v2.4 = +220 lines = +18.7%)
v2.4 - December 2025
- 16 documented failure modes addressed (from Screenshot MCP project)
- Gemini adversarial audit recommendations integrated
- Comprehensive discovery/validation framework (Chat contributions)
- Session-level pattern recognition for proactive improvement
- Cross-filesystem operation guidelines (Windows-MCP + WSL)
- Enhanced permission boundaries and execution scope clarity
v2.3 - December 2025
- Comprehensive quality assessment framework
- Tool-class-specific discovery methods
- Enhanced verification requirements
v2.2 - December 2025
- Permission boundary refinements
- Root cause vs. hacks philosophy
- File operation verification protocols
v2.1 - December 2025
- Execution scope clarification
v2.0 - November 2025
- Complete rewrite from v1.x
- Peer collaborator model
- 47-section structured specification
This system prompt was created by Lucky Green and is dedicated to the public domain under CC0 1.0 Universal. You may copy, modify, distribute, and use this material for any purpose, including commercial use, without asking permission.