research

name	research
description	Use when the user needs to research a technical topic, compare approaches, or build an evidence-based document with industry sources and expert opinions

Research Skill

Overview

Research a technical topic and produce a well-sourced document with direct quotes, industry references, and a clear recommendation.

When to Use

• Comparing architectural approaches (monorepo vs multi-repo, REST vs gRPC, etc.)
• Evaluating tools or technologies for adoption
• Building evidence-based ADRs or decision documents
• Answering "what do experts recommend for X?" questions
• Any time the user needs citations and industry backing for a decision

Input

The user provides a topic or question, optionally with:

• Context about their team/project
• Specific concerns to address
• Where to save the output

If not specified, save to docs/ in the current repo.

Workflow

digraph research_flow {
    rankdir=TB;
    node [shape=box];

    input [label="1. Clarify Topic & Scope"];
    research [label="2. Research (Parallel Agents)"];
    synthesize [label="3. Synthesize Findings"];
    write [label="4. Write Document"];
    review [label="5. Show Summary to User"];

    input -> research;
    research -> synthesize;
    synthesize -> write;
    write -> review;
}

Phase 1: Clarify Topic & Scope

Ask the user (if not already clear):

• What decision or question are they trying to answer?
• Who is the audience? (teammates, leadership, future self)
• Any constraints or preferences to bias toward?
• Where should the doc be saved?

Skip if the user already provided enough context.

Phase 2: Research (Use Parallel Agents)

Launch parallel Task agents to maximize coverage:

• Agent 1: Official sources — Documentation from the primary tools/frameworks involved (e.g., Terraform docs, Atmos docs, AWS docs)
• Agent 2: Industry voices — Blog posts, conference talks, and recommendations from recognized companies and engineers
• Agent 3: Community experience — GitHub discussions, Stack Overflow, Reddit, real-world case studies and post-mortems

Each agent should return:

• Direct quotes with attribution
• Source URLs
• Key data points (scale thresholds, timelines, costs)

Phase 3: Synthesize Findings

Combine research into a coherent narrative:

• Identify consensus across sources
• Note disagreements and what drives them (scale, team size, use case)
• Map findings to the user's specific context
• Form a clear recommendation with reasoning

Phase 4: Write Document

Use this structure:

# [Decision Title]

**Status:** Proposed
**Date:** [date]
**Context:** [1-2 sentences on what prompted this]

## Decision

[Clear statement of the recommended approach]

## Why Not [Alternative]?

[Side-by-side comparison table if applicable]

## What the Industry Recommends

### [Source 1 Name] ([Role/Context])
> *"Direct quote"*
> — [Source with link]

### [Source 2 Name] ([Role/Context])
> *"Direct quote"*
> — [Source with link]

[Repeat for 4-8 authoritative sources]

## How This Works in Practice

[Concrete examples, config snippets, or workflow descriptions relevant to the user's setup]

## Migration Path / Reversibility

[How to change course if the decision turns out wrong]

## When to Reconsider

| Signal | Threshold |
|--------|-----------|
| [trigger] | [specific number or condition] |

## Summary

[2-3 sentence wrap-up tying recommendation to evidence]

## References

- [All source URLs as clickable links]

Rules

1. Every claim needs a source — No unsourced opinions. If you can't find a source, say "based on general industry practice" rather than presenting it as cited fact.
2. Use direct quotes — Don't paraphrase when the original wording is impactful. Quote then explain.
3. Bias toward the user's context — A recommendation for a Fortune 500 is different from one for a 5-person startup. Always frame advice relative to the user's scale and situation.
4. Include dissenting views — If credible sources disagree, present both sides and explain what factors drive the disagreement.
5. Make it actionable — End with concrete next steps, not just "it depends."
6. Keep it skimmable — Use tables, headers, and bullet points. Teammates should get the gist in 2 minutes.