research-agent_skill

Overview

This skill runs open-ended research on any topic and builds a living Markdown research document. It supports an interactive mode for conversational exploration and a deep async mode for comprehensive investigations that run in the background. The document folder stores prompts, findings, exports, and related files so the research is reproducible and discoverable.

How this skill works

Interactive research creates a topic folder with prompt.md and research.md, then updates the document as we search, synthesize, and add sources in real time. Deep research uses an external parallel research processor to run multi-worker investigations, returns detailed Markdown output, and saves results into the same research folder. Every exchange updates the doc, notes confidence, and records sources and next steps.

When to use it

• Exploring an idea before committing to build or invest
• Investigating technical approaches, patterns, or trade-offs
• Running a market or competitive analysis that needs broad coverage
• Creating an ongoing living document for a research topic
• Scheduling long-running, parallelized research tasks for deep dives

Best practices

• Start each topic with prompt.md that captures the core question and start date
• Keep research.md structured: Open Questions, Findings, Options, Resources, Next Steps
• Record sources and confidence levels for every finding; use atomic bullets
• Run synthesis checkpoints every 5–10 exchanges to summarize and prune
• Save all outputs in the research folder and export PDFs into that folder

Example use cases

• Interactive exploration of a new open-source library with live summaries and code links
• Deep market scan for competitor features, pricing, and positioning over hours
• Technical deep-dive that aggregates documentation, benchmarks, and design notes
• Creating a reproducible research artifact for handoff to a project spec
• Archiving an evolving investigation as a dated reference with PDF export