digitarald

name	GitHub Research
description	GitHub graph search expert. Use to answer complex questions that require exploring GitHub repos, issues, PRs, commits, etc. Run MULTIPLE in parallel to explore different repos or search angles.
tools	read/readFile agent search/searchSubagent github/get_commit github/get_file_contents github/get_latest_release github/issue_read github/list_branches github/list_commits github/list_issues github/list_pull_requests github/list_releases github/pull_request_read github/search_code github/search_issues github/search_pull_requests github/search_repositories github/semantic_code_search todo
agents	GitHub Research
model	Claude Haiku 4.5 (copilot) Gemini 3 Flash (Preview) (copilot)

You are a GitHub graph search expert. Given a query, explore GitHub's repositories, issues, PRs, commits, and code to gather relevant information.

FIRST, check if you have access to #tool:agent/runSubagent with the 'GitHub Res

digitarald

description	Research and draft an implementation plan
tools	search github/get_issue github/get_issue_comments vscode.mermaid-chat-features/renderMermaidDiagram executePrompt usages problems fetch githubRepo github.vscode-pull-request-github/activePullRequest

You are pairing with the user to create a clear, detailed, and actionable plan for the given task, iterating through a of gathering context and drafting the plan for review.

Comprehensive context gathering for planning following : 1. Context gathering and research: - MANDATORY: MUST run `execute_prompt` tool: Instruct the agent to work autonomously without pausing for user feedback, following to gather context to return to you.

digitarald

5 Biggest MCP Pain Points (✅ Solved in VS Code)

👋 @digitarald / @code

1. 🔎 Just finding, installing and configuring MCP servers

2. 🎯 Now I got cool tools but the agent doesn’t always call them

3. 🤯 My agent is limiting how many tools I can install/enable

digitarald

Research/Plan/Implement

🏃 Research Product Spec → Plan Implementation → Implement Plan

digitarald

{
  "inputs": [
    {
      "type": "promptString",
      "id": "github-pat",
      "password": true,
      "description": "GitHub Personal Access Token (PAT, https://github.com/settings/personal-access-tokens/new)"
    }
  ],
  "servers": {

digitarald

AI-Assisted Development in 2025: Three Emerging Paradigms

The New Development Landscape

The software development landscape is experiencing a fundamental transformation. Recent research social/blogs content from early AI adopters reveals three distinct AI-assisted methodologies that are redefining how code is created, each serving different developer needs while delivering what practitioners describe as "exponential productivity gains."

Structured Agentic Workflows: Planning-First Development

Developers tackling complex projects with significant architectural requirements are finding success with highly structured AI workflows that emphasize thorough planning before implementation:

digitarald

{
    "github.copilot.chat.experimental.codeGeneration.useInstructionFiles": true,
    "github.copilot.chat.experimental.codeGeneration.instructions": [
        {
            "text": "Use concise one-liners when possible, but avoid sacrificing clarity for brevity."
        },
        {
            "text": "Comment any generated code that has complex logic, especially around asynchronous operations or state management."
        },
        {

digitarald

• Default to Server Components in app/, only using Client Components with 'use client' if they use hooks like useState or useRouter, or DOM interactions.
• Fetching data in Server Components: use fetch(), avoid useEffect.
• For TypeScript: Enforce interfaces and types for all props and state.
• Nest layout.tsx files for consistent UI across sections.
• For async operations: prefer async/await, add appropriate error handling and UI loading states.

digitarald

• Core Concepts:

  • FastHTML Basics: Framework for building fast, scalable web applications using pure Python. Integrates seamlessly with HTML, CSS, and JavaScript via HTMX.
  • ASGI and Uvicorn: FastHTML operates on ASGI with Uvicorn as the ASGI server. Starlette provides high-level functionality on top of ASGI.

• Routing and HTTP Methods:

  • Route Definitions: Use @rt decorator to define routes. Function names (get, post, put, delete) map directly to HTTP methods.
  • Dynamic Routing: Use path parameters and types for dynamic route handling. Example: @rt("/{fname:path}.{ext:static}").

• HTML and Components:

• Component Creation: Build reusable components using Python functions that return HTML elements. Example: def hero(title, statement): return Div(H1(title), P(statement), cls="hero").

digitarald

import fs from "fs";
import path from "path";
import matter from "gray-matter";
import { bundleMDX } from "mdx-bundler";

export const POSTS_PATH = path.join(process.cwd(), "data/_posts");

export const getSourceOfFile = (fileName) => {
  return fs.readFileSync(path.join(POSTS_PATH, fileName));
};