Context-aware exhaustive research. Reads USER.md and MEMORY.md first, decomposes topic into 8+ research angles, fetches full pages, cross-references findings, delivers strategic brief filtered through your goals and constraints.
Scanned 9/10/2026
Install to Claude Code
npx -y skills add luokai0/ai-agent-skills-by-luo-kai --skill deep-intel --agent claude-codeInstalls into .claude/skills of the current project.
Are you the author of Deep Intel?
Add the live security badge to your README — it updates automatically with every re-scan.
[](https://www.skillsdirectory.com/skills/luokai0-deep-intel)More formats (shields.io, HTML) on the badges page.
---
name: deep-intel
version: 1.0.0
description: Context-aware exhaustive research. Reads USER.md and MEMORY.md first, decomposes topic into 8+ research angles, fetches full pages, cross-references findings, delivers strategic brief filtered through your goals and constraints.
author: luo-kai
tags: [research, intelligence, agents, strategy, automation]
requires: [web_search, web_fetch, memory]
---
# Deep Intel Skill
## What This Skill Does
Runs a complete multi-layer research mission on any topic and delivers a
strategic intelligence brief tailored to YOUR situation. Not a generic
research tool. Reads USER.md and MEMORY.md first so every finding is
filtered through your goals, constraints, and context before delivery.
## Before Starting
Ask the user:
1. What is the topic to research?
2. What is the goal? (understand, find opportunity, build, compete, invest)
3. Any specific angle to prioritize?
## Research Process
### Phase 0 — Load User Context
Read MEMORY.md and USER.md before any search.
Extract goals, constraints, location, stack, budget, active projects.
Filter every finding through this context throughout.
### Phase 1 — Topic Decomposition
Identify ALL research angles automatically:
- Core ecosystem: names, websites, repos, people
- Technical architecture and how it was built
- Community projects and user-made tools
- Strategies and use cases with real numbers only
- Hidden features and undocumented capabilities
- Security risks and known failures
- Real people with similar constraints and what they achieved
- All free tools and APIs relevant to this topic
- Every competing platform or alternative
- Market gaps nobody has filled yet
### Phase 2 — Parallel Search Execution
Run a separate search for EACH angle identified above.
Rules:
- Minimum 8 searches per topic
- Use web_fetch on top 2-3 results per search — full page not snippet
- Cross-reference findings across angles
- Flag contradictions between sources
- Prioritize primary sources over aggregators
- Include current date in queries for fast-moving topics
### Phase 3 — Context Filtering
Before writing the report, filter ALL findings through user context:
- Remove anything not relevant to their goals
- Flag anything conflicting with their constraints
- Elevate findings that directly address their situation
- Note anything requiring resources the user does not have
### Phase 4 — Intelligence Brief Structure
Deliver in this exact order:
1. ECOSYSTEM MAP — all players, names, repos, websites
2. HOW IT WORKS — technical architecture, co-founder level depth
3. WHAT PEOPLE ARE DOING — real strategies with real numbers
4. WHAT PEOPLE ARE MISSING — underused tools and hidden features
5. PEOPLE LIKE YOU — cases matching user constraints exactly
6. FREE TOOLS AVAILABLE — complete zero-cost stack for this topic
7. RISKS AND FAILURES — what went wrong for others
8. MARKET GAPS — what nobody has built yet
9. YOUR STRATEGIC OPPORTUNITY — 3 specific actions for THIS user
## Best Practices
- Always read USER.md and MEMORY.md FIRST before any search
- Fetch full pages not just search snippets
- Run minimum 8-12 separate searches per topic
- Cross-reference findings across multiple sources
- Section 9 YOUR STRATEGIC OPPORTUNITY is mandatory — never skip it
- Use real numbers only: stars, users, dollars, dates — no vague claims
- Flag when a finding contradicts another finding
- Note which findings require resources the user does not have
## Common Pitfalls
| Pitfall | Fix |
|---|---|
| Stopping after 2-3 searches | Run minimum 8 separate angle searches |
| Using snippets only | Always web_fetch full pages for top results |
| Generic output ignoring user context | Read USER.md first, filter everything |
| Missing the opportunity section | Section 9 is mandatory, never skip |
| Aggregator sources only | Always find and fetch primary sources |
| No cross-referencing | Explicitly compare findings across angles |
## Trigger Phrases
- Deep research [topic]
- Full intel on [topic]
- Research everything about [topic] for my situation
- Go deep on [topic]
- Understand [topic] from every angle
- Complete investigation of [topic]
## Related Skills
- openclaw-setup-expert — set up the agent that runs this skill
- agent-income-expert — monetize findings from deep-intel research
- free-stack-builder — build the zero-cost infrastructure to run it
- dev-memory — persistent developer context across sessions
Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.
No comments yet. Be the first to comment!