/idea — Natural Language Idea Analysis

Analyze a natural language idea against the current project's codebase, assess feasibility, scope, and return structured JSON for issue creation.

Usage

/idea Add OAuth login with Google and GitHub providers
/idea 디스코드 봇에 모니터링 대시보드 추가

Workflow

Phase 1: Parse Input

Extract the core intent from the natural language input.

Phase 2: Codebase Analysis

1. Glob and Grep to understand current project structure
2. Identify relevant modules, files, and patterns
3. Check for existing similar functionality or conflicts

Phase 3: Feasibility Analysis

Analyze using a sonnet agent (always spawn with mode: "bypassPermissions"):

• Scope: Which modules/files would be affected
• Complexity: XS/S/M/L effort estimate
• Dependencies: What existing code needs to change
• Risks: Potential breaking changes or conflicts

Phase 4: Output

Return a JSON block with the analysis:

{
  "title": "concise feature title",
  "scope": "affected modules and files",
  "estimatedIssues": 3,
  "details": "2-3 sentence feasibility analysis",
  "issueSpecs": [
    {
      "title": "issue title",
      "body": "issue description with acceptance criteria",
      "labels": ["enhancement"]
    }
  ]
}

The JSON block MUST be wrapped in triple backtick json fence for parsing.

Model Selection

Phase	Model	Rationale
Phase 1-2	orchestrator	Simple read/search
Phase 3	sonnet	Balanced analysis

Integration

Rule	How
R009	Single agent for analysis
R010	Orchestrator manages phases; analysis delegated to agent with mode: "bypassPermissions"

Output Format

The output MUST contain a fenced JSON block that can be parsed by the calling system (e.g., builder-factory Discord bot). The JSON structure is the contract — do not change field names.