ToDoList.md

TODO List – Linter Configuration & Skill System Improvement

1. Multi-Platform Deployment

• Deploy the system to major commercial LLM platforms

  • OpenAI
  • Claude
  • Google (Gemini / PaLM)

• Deploy to major open-source LLM platforms

  • LLaMA-based platforms
  • Mistral / Mixtral
  • Other popular open-source serving frameworks

• Validate prompt compatibility and response consistency across platforms

---

2. Crawl and Prepare Linter Knowledge Base

• Select 5 major programming languages

  • Finalize language list

• For each language, select 3–5 popular linters

  • Validate popularity and ecosystem support
  • Check version compatibility

• Crawl and collect linter-related data

  • Official documentation
  • Configuration examples
  • Common rules and best practices

• Normalize and structure crawled data for model usage

---

3. Functional Testing with Diverse Prompts

• Test current functionality using different prompt styles

3.1 Task Description Variants

• "generate linter configuration"
• "how to enforce XXXX"
• "configure XXXXX"

3.2 Linter Specification

• Prompt explicitly specifies the linter
• Prompt does NOT specify the linter

3.3 Crawled Data Coverage

• Linter exists in crawled data
• Linter does NOT exist in crawled data

3.4 Coding Standard Input Variations

• Coding standard provided as plain string

• Coding standard provided via file path

• Coding standard provided via URL

• Coding standard length variations

  • Single sentence
  • No explicit coding standard
  • One coding rule
  • Two coding rules
  • Two sentences that represent a single coding rule

3.5 Code Input Variations

• No code provided
• Code provided as string
• Code provided via file path
• Code provided via URL

---

4. Skill.md Design Improvements

• Review current Skill.md design (prompt-driven only)

• Identify gaps between prompt logic and actual code execution

• Improve integration between:

  • Prompt design
  • Code execution flow
  • Linter invocation and result handling

• Redesign Skill.md to better reflect:

  • End-to-end workflow
  • Prompt + code collaboration
  • Extensibility for new linters and languages

• Validate the new Skill.md with real-world use cases