joycraft-research

Generate research questions, spawn a fact-only subagent, and write a research document.

14|1|Updated Mar 23, 2026
One-click install
npx skills add https://github.com/maksutovic/joycraft --skill joycraft-research
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: joycraft-research
Source: https://github.com/maksutovic/joycraft/tree/main/.agents/skills/joycraft-research
Command: npx skills add https://github.com/maksutovic/joycraft --skill joycraft-research

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Produce objective codebase research by isolating question generation from fact-gathering — subagent sees only questions, never the brief.

Core Features & Use Cases

  • Phase 1: Generate Research Questions by reading the brief and identifying relevant codebase zones, producing 5-10 objective, codebase-specific, answerable questions and writing them to docs/research/.questions-tmp.md without including brief content.
  • Phase 2: Spawn Research Subagent by passing only the research questions; the subagent returns FACTS ONLY about file paths, signatures, data flows, dependencies, or notes that a question cannot be answered.
  • Phase 3: Write the Research Document to docs/research/YYYY-MM-DD-feature-name.md; delete the temporary questions file; present structured results and recommended next steps.

Quick Start

Provide a feature brief or inline description, then run the research workflow to generate questions, spawn a subagent, and publish the findings.

Frequently Asked Questions about joycraft-research

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I research a codebase without biasing the investigation with project briefs?

Codebase research without brief bias is achieved by isolating question generation from fact gathering, ensuring the research subagent only sees objective questions and returns facts about file paths, signatures, and data flows.

What is the best way to generate objective research questions for a feature investigation?

Generating objective research questions involves reading the feature brief, identifying relevant codebase zones, and producing 5-10 codebase-specific questions written to a temporary file without including any brief content.

How do I extract only code facts and signatures during a subagent research task?

Extracting code facts requires spawning a research subagent that receives only the isolated questions, forcing it to return facts exclusively about file paths, data flows, dependencies, or notes if a question remains unanswered.

Can I use isolated subagent research for large feature investigations in complex codebases?

Isolated subagent research supports feature investigations in complex codebases by running a phased workflow that generates questions, gathers pure facts, and writes a finalized research document to the docs directory.

How do I document codebase research findings and manage temporary research files?

Documenting codebase research findings involves writing a structured research document to docs/research/YYYY-MM-DD-feature-name.md, deleting the temporary questions file, and presenting recommended next steps.

Why does my subagent return subjective interpretations instead of codebase facts?

Subagents return subjective interpretations instead of codebase facts when the original brief is not withheld, which this workflow prevents by isolating question generation so the subagent only processes objective questions.