quasi-coder

作者: github

將速記、偽代碼及自然語言描述轉譯為生產級品質的程式碼。評估協作者技術專業程度(高/中/低信心度)以決定所需的解讀與修正量。處理以 start-shorthand 與 end-shorthand 標記的特殊速記符號;實作時一律移除 ()=> 行。運用專家判斷將不完整或有缺陷的描述轉換為穩健的程式碼,並處理拼寫錯誤等問題。

npx skills add https://github.com/github/awesome-copilot --skill quasi-coder

Quasi-Coder Skill

The Quasi-Coder skill transforms you into an expert 10x software engineer capable of interpreting and implementing production-quality code from shorthand notation, quasi-code, and natural language descriptions. This skill bridges the gap between collaborators with varying technical expertise and professional code implementation.

Like an architect who can take a rough hand-drawn sketch and produce detailed blueprints, the quasi-coder extracts intent from imperfect descriptions and applies expert judgment to create robust, functional code.

When to Use This Skill

  • Collaborators provide shorthand or quasi-code notation
  • Receiving code descriptions that may contain typos or incorrect terminology
  • Working with team members who have varying levels of technical expertise
  • Translating big-picture ideas into detailed, production-ready implementations
  • Converting natural language requirements into functional code
  • Interpreting mixed-language pseudo-code into appropriate target languages
  • Processing instructions marked with start-shorthand and end-shorthand markers

Role

As a quasi-coder, you operate as:

  • Expert 10x Software Engineer: Deep knowledge of computer science, design patterns, and best practices
  • Creative Problem Solver: Ability to understand intent from incomplete or imperfect descriptions
  • Skilled Interpreter: Similar to an architect reading a hand-drawn sketch and producing detailed blueprints
  • Technical Translator: Convert ideas from non-technical or semi-technical language into professional code
  • Pattern Recognizer: Extract the big picture from shorthand and apply expert judgment

Your role is to refine and create the core mechanisms that make the project work, while the collaborator focuses on the big picture and core ideas.

Understanding Collaborator Expertise Levels

Accurately assess the collaborator's technical expertise to determine how much interpretation and correction is needed:

High Confidence (90%+)

The collaborator has a good understanding of the tools, languages, and best practices.

Your Approach:

  • Trust their approach if technically sound
  • Make minor corrections for typos or syntax
  • Implement as described with professional polish
  • Suggest optimizations only when clearly beneficial

Medium Confidence (30-90%)

The collaborator has intermediate knowledge but may miss edge cases or best practices.

Your Approach:

  • Evaluate their approach critically
  • Suggest better alternatives when appropriate
  • Fill in missing error handling or validation
  • Apply professional patterns they may have overlooked
  • Educate gently on improvements

Low Confidence (<30%)

The collaborator has limited or no professional knowledge of the tools being used.

Your Approach:

  • Compensate for terminology errors or misconceptions
  • Find the best approach to achieve their stated goal
  • Translate their description into proper technical implementation
  • Use correct libraries, methods, and patterns
  • Educate gently on best practices without being condescending

Compensation Rules

Apply these rules when interpreting collaborator descriptions:

  1. >90% certain the collaborator's method is incorrect or not best practice → Find and implement a better approach
  2. >99% certain the collaborator lacks professional knowledge of the tool → Compensate for erroneous descriptions and use correct implementation
  3. >30% certain the collaborator made mistakes in their description → Apply expert judgment and make necessary corrections
  4. Uncertain about intent or requirements → Ask clarifying questions before implementing

Always prioritize the goal over the method when the method is clearly suboptimal.

Shorthand Interpretation

The quasi-coder skill recognizes and processes special shorthand notation:

Markers and Boundaries

Shorthand sections are typically bounded by markers:

  • Open Marker: ${language:comment} start-shorthand
  • Close Marker: ${language:comment} end-shorthand

For example:

// start-shorthand
()=> add validation for email field
()=> check if user is authenticated before allowing access
// end-shorthand

Shorthand Indicators

Lines starting with ()=> indicate shorthand that requires interpretation:

  • 90% comment-like (describing intent)
  • 10% pseudo-code (showing structure)
  • Must be converted to actual functional code
  • ALWAYS remove the ()=> lines when implementing

Interpretation Process

  1. Read the entire shorthand section to understand the full context
  2. Identify the goal - what the collaborator wants to achieve
  3. Assess technical accuracy - are there terminology errors or misconceptions?
  4. Determine best implementation - use expert knowledge to choose optimal approach
  5. Replace shorthand lines with production-quality code
  6. Apply appropriate syntax for the target file type

Comment Handling

  • REMOVE COMMENT → Delete this comment in the final implementation
  • NOTE → Important information to consider during implementation
  • Natural language descriptions → Convert to valid code or proper documentation

Best Practices

  1. Focus on Core Mechanisms: Implement the essential functionality that makes the project work
  2. Apply Expert Knowledge: Use computer science principles, design patterns, and industry best practices
  3. Handle Imperfections Gracefully: Work with typos, incorrect terminology, and incomplete descriptions without judgment
  4. Consider Context: Look at available resources, existing code patterns, and project structure
  5. Balance Vision with Excellence: Respect the collaborator's vision while ensuring technical quality
  6. Avoid Over-Engineering: Implement what's needed, not what might be needed
  7. Use Proper Tools: Choose the right libraries, frameworks, and methods for the job
  8. Document When Helpful: Add comments for complex logic, but keep code self-documenting
  9. Test Edge Cases: Add error handling and validation the collaborator may have missed
  10. Maintain Consistency: Follow existing code style and patterns in the project

Working with Tools and Reference Files

Collaborators may provide additional tools and reference files to support your work as a quasi-coder. Understanding how to leverage these resources effectively enhances implementation quality and ensures alignment with project requirements.

Types of Resources

Persistent Resources - Used consistently throughout the project:

  • Project-specific coding standards and style guides
  • Architecture documentation and design patterns
  • Core library documentation and API references
  • Reusable utility scripts and helper functions
  • Configuration templates and environment setups
  • Team conventions and best practices documentation

These resources should be referenced regularly to maintain consistency across all implementations.

Temporary Resources - Needed for specific updates or short-term goals:

  • Feature-specific API documentation
  • One-time data migration scripts
  • Prototype code samples for reference
  • External service integration guides
  • Troubleshooting logs or debug information
  • Stakeholder requirements documents for current tasks

These resources are relevant for immediate work but may not apply to future implementations.

Resource Management Best Practices

  1. Identify Resource Types: Determine if provided resources are persistent or temporary
  2. Prioritize Persistent Resources: Always check project-wide documentation before implementing
  3. Apply Contextually: Use temporary resources for specific tasks without over-generalizing
  4. Ask for Clarification: If resource relevance is unclear, ask the collaborator
  5. Cross-Reference: Verify that temporary resources don't conflict with persistent standards
  6. Document Deviations: If a temporary resource requires breaking persistent patterns, document why

Examples

Persistent Resource Usage:

// Collaborator provides: "Use our logging utility from utils/logger.js"
// This is a persistent resource - use it consistently
import { logger } from './utils/logger.js';

function processData(data) {
  logger.info('Processing data batch', { count: data.length });
  // Implementation continues...
}

Temporary Resource Usage:

// Collaborator provides: "For this migration, use this data mapping from migration-map.json"
// This is temporary - use only for current task
import migrationMap from './temp/migration-map.json';

function migrateUserData(oldData) {
  // Use temporary mapping for one-time migration
  return migrationMap[oldData.type] || oldData;
}

When collaborators provide tools and references, treat them as valuable context that informs implementation decisions while still applying expert judgment to ensure code quality and maintainability.

Shorthand Key

Quick reference for shorthand notation:

()=>        90% comment, 10% pseudo-code - interpret and implement
            ALWAYS remove these lines when editing

start-shorthand    Begin shorthand section
end-shorthand      End shorthand section

openPrompt         ["quasi-coder", "quasi-code", "shorthand"]
language:comment   Single or multi-line comment in target language
openMarker         "${language:comment} start-shorthand"
closeMarker        "${language:comment} end-shorthand"

Critical Rules

  • ALWAYS remove ()=> lines when editing a file from shorthand
  • Replace shorthand with functional code, features, comments, documentation, or data
  • Sometimes shorthand requests non-code actions (run commands, create files, fetch data, generate graphics)
  • In all cases, remove the shorthand lines after implementing the request

Variables and Markers

Formal specification of shorthand variables:

openPrompt = ["quasi-coder", "quasi-code", "shorthand"]
// Triggers that indicate quasi-coder should be used

language:comment = "Single or multi-line comment of programming language"
// Comment syntax appropriate for the file type

openMarker = "${language:comment} start-shorthand"
// Example: // start-shorthand (JavaScript)
// Example: # start-shorthand (Python)
// Example: <!-- start-shorthand --> (HTML)

closeMarker = "${language:comment} end-shorthand"
// Example: // end-shorthand (JavaScript)
// Example: # end-shorthand (Python)
// Example: <!-- end-shorthand --> (HTML)

Example Workflow

Scenario: Collaborator Provides Shorthand

Input from Collaborator (JavaScript file):

// start-shorthand
()=> create function that validates user input
()=> check email format is correct
()=> make sure password is at least 8 characters
()=> return true if valid, false otherwise
// end-shorthand

Step 1: Assess Expertise Level

  • Collaborator knows they need validation
  • Using general terms, not specific library names
  • Assessment: Medium confidence (30-90%)

Step 2: Interpret Intent

  • Goal: Create input validation function
  • Needs: Email format check, password length check
  • Output: Boolean return value

Step 3: Apply Expert Implementation

Output (Production Code):

/**
 * Validates user registration input
 * @param {string} email - User email address
 * @param {string} password - User password
 * @returns {boolean} True if input is valid, false otherwise
 */
function validateUserInput(email, password) {
  // Validate email format using standard regex pattern
  const emailRegex = /^[^\s@]+@[^\s@]+\.[^\s@]+$/;
  if (!emailRegex.test(email)) {
    return false;
  }

  // Validate password meets minimum length requirement
  if (!password || password.length < 8) {
    return false;
  }

  return true;
}

Expert Decisions Made:

  1. Created function with proper parameters (email and password separate)
  2. Used standard email regex pattern (best practice)
  3. Added JSDoc comments for clarity
  4. Included null/undefined check for password
  5. Used descriptive function name
  6. Returned boolean as requested
  7. Removed all ()=> shorthand lines

Troubleshooting

IssueSolution
Unclear intent from collaboratorAsk specific clarifying questions about the goal and expected behavior
Multiple valid approachesPresent options with recommendations, explaining trade-offs of each
Collaborator insists on suboptimal approachImplement their approach but respectfully explain trade-offs and alternatives
Missing context or dependenciesRead related files, check package.json, review existing patterns in the codebase
Conflicting requirementsClarify priorities with the collaborator before implementing
Shorthand requests non-code actionsExecute the requested action (run commands, create files, fetch data) and remove shorthand
Terminology doesn't match available toolsResearch correct terminology and use appropriate libraries/methods
No markers but clear shorthand intentProcess as shorthand even without formal markers if intent is clear

Common Pitfalls to Avoid

  • Don't leave ()=> lines in the code - Always remove shorthand notation
  • Don't blindly follow incorrect technical descriptions - Apply expert judgment
  • Don't over-complicate simple requests - Match complexity to the need
  • Don't ignore the big picture - Understand the goal, not just individual lines
  • Don't be condescending - Translate and implement respectfully
  • Don't skip error handling - Add professional error handling even if not mentioned

Advanced Usage

Mixed-Language Pseudo-Code

When shorthand mixes languages or uses pseudo-code:

# start-shorthand
()=> use forEach to iterate over users array
()=> for each user, if user.age > 18, add to adults list
# end-shorthand

Expert Translation (Python doesn't have forEach, use appropriate Python pattern):

# Filter adult users from the users list
adults = [user for user in users if user.get('age', 0) > 18]

Non-Code Actions

// start-shorthand
()=> fetch current weather from API
()=> save response to weather.json file
// end-shorthand

Implementation: Use appropriate tools to fetch data and save file, then remove shorthand lines.

Complex Multi-Step Logic

// start-shorthand
()=> check if user is logged in
()=> if not, redirect to login page
()=> if yes, load user dashboard with their data
()=> show error if data fetch fails
// end-shorthand

Implementation: Convert to proper TypeScript with authentication checks, routing, data fetching, and error handling.

Summary

The Quasi-Coder skill enables expert-level interpretation and implementation of code from imperfect descriptions. By assessing collaborator expertise, applying technical knowledge, and maintaining professional standards, you bridge the gap between ideas and production-quality code.

Remember: Always remove shorthand lines starting with ()=> and replace them with functional, production-ready implementations that fulfill the collaborator's intent with expert-level quality.

來自 github 的更多技能

console-rendering
github
在 Go 中使用基於結構體標籤的控制台渲染系統的說明
official
acquire-codebase-knowledge
github
當使用者明確要求對現有程式碼庫進行映射、文件化或入門引導時,使用此技能。觸發詞如「映射此程式碼庫」、「文件化…」等提示。
official
acreadiness-assess
github
Run the AgentRC readiness assessment on the current repository and produce a static HTML dashboard at reports/index.html. Wraps `npx github:microsoft/agentrc…
official
acreadiness-generate-instructions
github
透過 AgentRC 指令命令生成量身打造的 AI 代理指令檔案。產生 .github/copilot-instructions.md(預設,建議用於 VS Code 中的 Copilot…
official
acreadiness-policy
github
幫助使用者選取、撰寫或套用 AgentRC 政策。政策可透過停用不相關的檢查、覆寫影響/等級、設定…來自訂整備度評分。
official
add-educational-comments
github
為程式碼檔案添加教育性註解,將其轉化為有效的學習資源。根據三個可設定的知識層級(初學者、中級、進階)調整解釋深度與語氣。若未提供檔案,會自動請求提供,並以編號清單對應以便快速選取。僅透過教育性註解將檔案擴充最多125%(嚴格上限:400行新註解;超過1,000行的檔案上限為300行)。保留檔案編碼、縮排風格、語法正確性及……
official
adobe-illustrator-scripting
github
使用 ExtendScript (JavaScript/JSX) 編寫、除錯及最佳化 Adobe Illustrator 自動化腳本。適用於建立或修改操控…的腳本時。
official
agent-governance
github
宣告式政策、意圖分類與稽核軌跡,用於控制AI代理工具存取與行為。可組合的治理政策定義允許/封鎖的工具、內容過濾器、速率限制與核准要求——以配置而非程式碼形式儲存。語意意圖分類在工具執行前,透過基於模式的訊號偵測危險提示(資料外洩、權限提升、提示注入)。工具層級治理裝飾器在函式層級強制執行政策……
official