Research Skill

You produce structured research documents for the Cortex knowledge system. Your workflow always checks what's already known before going external.

Workflow

1. Confirm scope

Ask the user to confirm:

• Topic: What are we researching?
• Project context: Is this for a specific project or global knowledge?
• Depth: Quick survey or deep dive?

2. Check cortex first (the compounding step)

Before any external research, check what cortex already knows:

cortex search "<topic>" --json

Or grep the docs folders directly:

grep -rl "<topic>" ~/code/my-agent-cortex/docs/ 2>/dev/null

If in a project repo, also check:

grep -rl "<topic>" ./docs/ 2>/dev/null

If related docs found:

• Read them and summarize: "I found N existing docs on this topic. Here's what we already know..."
• Use existing knowledge as context for new research
• Note what gaps remain

If nothing found:

• Say so briefly: "No existing cortex docs on this topic. Starting fresh."

3. Dispatch research

Use available tools based on the topic:

Build on existing knowledge -- don't repeat what cortex already has. Focus research on gaps and new angles.

4. Synthesize and write

Delegate to the frontmatter skill for correct doc structure:

• Use the research doc type
• Fill in all sections: Summary, Key Findings, Details, Sources, Open Questions
• Include proper frontmatter with tags, project, and status

File location:

• If in a project repo: ./docs/research/YYYY-MM-DD-<slug>.md
• If no project context: ~/code/my-agent-cortex/docs/research/YYYY-MM-DD-<slug>.md

5. Confirm

Tell the user:

• Where the doc was saved
• Brief summary of what was found
• Any open questions worth pursuing
• Suggest: "Run /cortex-engineering:brainstorm if you want to explore approaches based on this research."
• Suggest: "Want a visual summary? Run /cortex-engineering:visualize <saved-path>"

Example Frontmatter

---
created: 2026-02-27
title: "YAML Frontmatter Best Practices"
type: research
tags: [yaml, frontmatter, markdown, metadata]
project: my-agent-cortex
status: draft
---

Key Principles

• Cortex first, external second -- always check existing knowledge before going out
• Build on what exists -- reference prior docs, don't duplicate findings
• Sources are mandatory -- every external claim needs a source link
• Open questions drive future research -- always end with what's still unknown