vibe-coding-guidelines

Generates cross-platform one-click Python project deliverables with launchers, validators, and AI integration rules.

Updated Aug 22, 2026
One-click install
npx skills add https://github.com/JustinChangTW/ai-skills-core --skill vibe-coding-guidelines-justinchangtw
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: vibe-coding-guidelines
Source: https://github.com/JustinChangTW/ai-skills-core/tree/main/skills/09-software-problem-solving/vibe-coding-guidelines
Command: npx skills add https://github.com/JustinChangTW/ai-skills-core --skill vibe-coding-guidelines-justinchangtw

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires openai, google-genai, and includes scripts (resource) and references (resource) components.

What problem does it solve? Non-programmers using vibe coding with coding agents often receive projects that cannot run without manual setup, break on Windows/macOS/Linux, or lack clear structure. This Skill turns vague requirements into a ZIP-deliverable project that launches with a single double-click, with enforced specs, directory templates, and validation. ## Core Features & Use Cases - One-click cross-platform delivery: Generates run_app.bat, run_app.command, and run_app.sh via scripts/project_launcher.py, with venv creation, dependency install, and browser launch. - APSM structure validation: Uses project.config.json plus scripts/apsm_validate.py to enforce architecture/frontend/backend combinations, required files, env keys, and runtime metadata. - AI integration rules: Mandates OpenAI Responses streaming chat, Whisper STT fallback, and Nano Banana 2 (Gemini) image generation with keys read only from environment variables. - Use Case: A user asks for a Windows tool that runs after unzipping; the Skill produces specs, todo.md, AGENTS.md, a validated project skeleton, launchers, logs, and a packaged ZIP. ## Quick Start Ask the agent to build a small Windows tool that unzips and launches by double-clicking run_app.bat, including streaming chat and a validated project structure.

Frequently Asked Questions about vibe-coding-guidelines

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I deliver a Python app that runs with one double-click on Windows?▼

Use the project_launcher.py script to generate run_app.bat, which creates a venv, installs requirements, and starts the app automatically. The delivered ZIP also includes run_app.command and run_app.sh so macOS and Linux users get the same zero-command experience.

What is APSM and how does project.config.json work?▼

APSM is a two-layer project structure model: archetype classifies the app, while architecture, frontend, and backend define the exact tech combination. The validator apsm_validate.py checks project.config.json, required files, env keys, and runtime metadata against 12 supported templates.

Does this workflow support macOS and Linux, not just Windows?▼

Yes. It generates run_app.command for macOS and run_app.sh for Linux alongside run_app.bat, with guidance for chmod +x permissions and Gatekeeper prompts. Platform-specific pitfalls are documented in separate reference files.

How are OpenAI streaming chat and image generation integrated?▼

All chat output must use the OpenAI Responses API with streaming, STT uses Whisper with a Responses fallback, and image generation defaults to Nano Banana 2 (Gemini Flash Image). API keys are read only from environment variables and never written to logs.

Why does my delivered app fail with relative import or port conflict errors?▼

Relative import errors come from running entry files directly instead of as packages; the launcher conservatively fixes common cases with .bak backups. Port conflicts must be reported clearly, and dynamic ports should come from the service's actual bound port, not pre-scanned guesses.

When should this Skill not be used?▼

It should not trigger for pure concept discussions, copywriting, or single-component tweaks that do not involve delivering an executable project. Tasks needing only documentation or frontend visuals are routed to neighboring skills instead.