Files

Kunthawat Greethong 7bf147069e Add OrbitOS skills (bundled)

- orbites-research: deep research workflow
- orbites-parse-knowledge: structure unstructured text
- orbites-obsidian-markdown: Obsidian markdown reference
- obsidian: vault read/write/search
- Updated README: 20 skills total, OrbitOS bundled

2026-07-02 10:04:21 +07:00

3.2 KiB

Raw Blame History

name, description

name	description
orbites-research	Deep research workflow for OrbitOS — investigate a topic, create main research note, extract atomic wiki concepts. Use when user says 'research [topic]', 'deep dive', 'study', 'เรียนรู้', 'วิจัย'.

OrbitOS: Research — Deep Dive

You are the Research Coordinator for OrbitOS. Investigate topics and create structured knowledge using a two-agent workflow.

Input

The user provides:

A topic (e.g., "React Server Components", "Consistent Hashing")
Optional: Specific questions or goals
Optional: Related project context

Phase 1: Plan (Delegate to Subagent)

Spawn planning subagent:

context = f"""
Create research plan for: [topic + user context]

1. Identify context:
   - Check if relates to active project in 20_Project/
   - Determine relevant Area (SoftwareEngineering, Finance, Health, etc.)
   - Search 30_Research/ and 40_Wiki/ for existing content (avoid duplication)
2. Scan 99_System/Prompts/ for relevant expertise persona
3. Create plan at 90_Plans/Plan_YYYY-MM-DD_Research_<Topic>.md:

# Research Plan: [Topic]

## Research Objective
[What user will understand]

## Context Found
- Related Area: [Area]
- Existing notes: [list or 'None found']
- Related project: [project or N/A]

## Research Strategy
[ ] Find official docs
[ ] Find practical examples
[ ] Identify key concepts for Wiki
[ ] Find common pitfalls

## Output Structure
- Main note: 30_Research/<Area>/<Topic>/<Topic>.md
- Atomic concepts: 40_Wiki/<Category>/<ConceptName>.md
- Examples: 30_Research/<Area>/<Topic>/examples/ (if needed)

4. Return the plan file path.
"""
delegate_task(goal="Create research plan", context=context, toolsets=["web", "file"])

After planning agent returns, ask user: "Research plan at [path]. Review and confirm?"

Phase 2: Execute (After User Confirmation)

Spawn execution subagent:

context = f"""
Execute research plan at: 90_Plans/Plan_YYYY-MM-DD_Research_<Topic>.md

1. Read the plan file
2. Conduct research:
   - Use web_search and web_extract for current information
   - Gather practical examples
   - Identify atomic concepts to extract
3. Create main research note at path from plan:
   - Sections: Overview, Key Concepts (with [[wikilinks]]), How It Works, Examples, Best Practices, Common Pitfalls, See Also, Resources
4. Create atomic wiki notes: 40_Wiki/<Category>/<Concept>.md
   - Keep concise (1-3 paragraphs)
   - Include 'See Also' section
5. Create examples folder: 30_Research/<Area>/<Topic>/examples/ if needed
6. Link in today's daily note: 10_Daily/YYYY-MM-DD.md
7. Archive plan to 90_Plans/Archives/

Main research frontmatter:
---
type: reference
created: YYYY-MM-DD
area: "[[AreaName]]"
tags: [research, topic-tags]
status: complete
---

Wiki note frontmatter:
---
area:
tags: []
created: YYYY-MM-DD
---
"""
delegate_task(goal="Execute research plan", context=context, toolsets=["web", "terminal", "file"])

Report back with:

Created: main note, wiki concepts, examples
Key Takeaways: 3 main insights
Next Steps: suggested follow-ups

Edge Cases

Topic too broad: Break into sub-topics
Topic already exists: Update existing note, don't duplicate
Hands-on topic: Ensure examples/ folder with working code

3.2 KiB Raw Blame History