The start prompt

/brainstorming i like to build a app that will install Copilot hock in a repo that the user points to and, it should make hock for all the states that are in Github copilot-cli, meaning i like to save all prompts and i like to save them i a Github repo under a Issue, so all prompts will be saved here in the repo but as comment to first isus, by that i will have all prompts made in the copilot-cli doing the vibe coding, under the same issue The app should the have a other function that is user points to a Github repo (maybe the one above) but any and also the issue, then the app should analyze the issue incl all the comments, and give a summary og how good are the prompts overall and on every single prompt, with a carector from 1 to 5 and a short comment on what could improve it and what is good
Build MiniMax M2.5 Free OpenCode Zen

It was good in asking questions! and i added information and needs until the plan was written, this is very good i am convinced that the better the plan is from start the faster and less fails you will have.

The plan it came back with (i will make a Post on how it ended up with, even if its go wrong)

I might try this also in Github copilot-cli with Superpower installed

---

Prompt Helper and Analyzer - Implementation Plan

Overview

A TypeScript/Node tool that captures Copilot CLI prompts via hooks and provides a web dashboard to analyze prompt quality using Copilot SDK.

Architecture

Components

1. CLI Tool - Generates hook configs, manages GitHub integration
2. Copilot CLI Hooks - Captures prompts via .github/hooks/copilot-journal.json
3. Web Dashboard - Uses Copilot SDK for prompt analysis and history viewing

Data Flow

User types prompt in Copilot CLI
    ↓
userPromptSubmitted hook fires
    ↓
Hook script posts prompt as GitHub issue comment
    ↓
Dashboard fetches issue + comments → analyzes with Copilot SDK

Commands

prompt-helper init [--issue-title "My Journal"]

1. Check gh authentication
2. Create GitHub issue in current repo
3. Generate .github/hooks/copilot-journal.json
4. Create hook scripts in .github/hooks/scripts/

prompt-helper analyze <issue-url> [--model gpt-5.1]

• Fetch issue and all comments
• Use Copilot SDK to analyze each prompt
• Output scores and feedback

prompt-helper models

• Fetch available models from Copilot CLI
• Display to user

prompt-helper dashboard [--port 3000]

• Start web server
• Serve dashboard UI

Configuration

.prompt-helper/config.json

jsonCopy
{
  "defaultModel": "gpt-5.1",
  "scoringWeights": {
    "clarity": 40,
    "context": 30,
    "completeness": 30
  },
  "github": {
    "tokenEnvVar": "GH_TOKEN"
  }
}

.prompt-helper/repos.json

jsonCopy
[
  {
    "repo": "owner/project",
    "issueNumber": 1,
    "issueTitle": "Copilot Journal - 2026-04-17"
  }
]

Scoring Criteria (Configurable)

Dimension	Weight (default)	Description
Clarity	40%	Is the intent clear? Specific files mentioned? Desired outcome stated?
Context	30%	Relevant background? Constraints? Scope appropriate?
Completeness	30%	All details included? Edge cases considered? Acceptance criteria clear?

Analysis Output (per prompt)

• Score: 1-5 stars
• What's good ✓
• What could improve ✗
• Overall weighted average

Project Structure

prompt-helper/
├── src/
│   ├── cli/
│   │   ├── commands/
│   │   │   ├── init.ts
│   │   │   ├── analyze.ts
│   │   │   ├── models.ts
│   │   │   └── dashboard.ts
│   │   └── index.ts
│   ├── hooks/
│   │   └── capture-prompt.ts
│   ├── lib/
│   │   ├── github.ts
│   │   ├── config.ts
│   │   └── analyzer.ts
│   └── dashboard/
│       ├── server.ts
│       └── public/
├── package.json
├── tsconfig.json
└── README.md

Hook Files Structure

<repo>/.github/hooks/
├── copilot-journal.json    # Hook config
└── scripts/
    ├── capture-prompt.sh   # Bash (Linux/Mac)
    └── capture-prompt.ps1  # PowerShell (Windows)

Dashboard Features

Home Page

• List tracked repos and issues
• Stats: total prompts, average score, sessions count

Issue Detail

• All prompts from issue as comments
• Each prompt: timestamp, cwd, full text
• "Analyze" button per prompt

Analysis Results

• Overall score (weighted average)
• Per-prompt scores with feedback

Export

• Download as Markdown
• Download as JSON

Settings

• Model selector (from Copilot CLI)
• Sliders for scoring weights (must sum to 100%)

Tech Stack

• TypeScript
• Node.js
• Express.js or Fastify
• @github/copilot-sdk
• GitHub API (via gh CLI or octokit)

Implementation Tasks

Phase 1: CLI Core

1. Initialize project with TypeScript
2. Set up CLI framework (commander or yargs)
3. Implement config management
4. Implement GitHub API integration (gh CLI)

Phase 2: Hook System

5. Create init command
6. Generate hook JSON config
7. Create hook scripts (bash + PowerShell)
8. Test hook integration

Phase 3: Analysis

9. Implement models command (fetch available models)
10. Integrate Copilot SDK
11. Implement analyze command
12. Implement scoring logic

Phase 4: Dashboard

13. Set up Express/Fastify server
14. Create dashboard UI
15. Implement issue fetching
16. Implement analysis in dashboard
17. Add export functionality
18. Add settings page

Phase 5: Polish

19. Add error handling
20. Write tests
21. Build and package
22. Documentation