research

Coordinate bounded parallel subagents to produce a unified research.md artifact.

4|Updated Apr 26, 2026
One-click install
npx skills add https://github.com/curdx/curdx-flow --skill research-curdx
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: research
Source: https://github.com/curdx/curdx-flow/tree/main/plugins/curdx-flow/skills/research
Command: npx skills add https://github.com/curdx/curdx-flow --skill research-curdx

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

This Skill prevents spec discovery from becoming fragmented by coordinating a repeatable research phase that produces a consolidated research artifact tied to the active spec.

Core Features & Use Cases

  • Spec-rooted research execution: Resolves the active spec, reads goal/state/progress, and drafts research inputs based on existing context.
  • Interactive interview planning (optional): Uses a dialogue framework to classify intent and avoid repeating already-answered questions.
  • Bounded parallel discovery: Dispatches multiple subagents for external-topic research and codebase exploration, then merges outputs into a single research.md.
  • Validation and approval gates: Optionally runs a spec-reviewer loop, presents a walkthrough, and waits for explicit user approval before finalizing.
  • Stateful workflow progress: Updates .curdx-state.json and .progress.md, optionally commits the artifact, and stops before proceeding to requirements.

Quick Start

Start a research run for the current active spec by invoking /curdx-flow:research <spec-name> and follow the approval prompt.

Frequently Asked Questions about research

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

FAQPage Schema
How do I coordinate parallel discovery research for an active spec?

To coordinate parallel discovery research, you invoke the Skill to resolve the active spec snapshot, dispatch bounded parallel subagents for codebase exploration, and merge their outputs into a unified research artifact.

What is a research artifact and when do I need one for spec reruns?

A research artifact is a consolidated research.md file capturing targeted findings and context. You need one during spec reruns to prevent fragmented discovery and ensure all research is unified before proceeding.

How do I validate research findings before moving to requirements?

You validate research findings by running an optional spec-reviewer loop, presenting a walkthrough of the merged research.md, and waiting for explicit user approval before finalizing the workflow state.

Can I use interactive interview planning to avoid repeating already-answered questions?

Yes, you can use the optional interactive interview planning feature, which uses a dialogue framework to classify intent and prevent asking questions that have already been answered in the active spec context.

What are the limitations of bounded parallel subagents for codebase exploration?

Bounded parallel subagents are limited to targeted findings within the codebase and external topics, requiring a strict merge step into a single artifact and safe state finalization that stops before requirements generation.

Does the research workflow require a snapshot-based spec resolution?

Yes, snapshot-based spec resolution is required to read the goal, state, and progress, ensuring the dispatched subagents gather context based on the exact active spec snapshot.