Deep Research — Claude Skill

A Claude skill that delivers 10-15x the depth of surface-level research on any topic — markets, products, features, competitors, regulations, or opportunities.

What This Skill Does

Most AI research outputs are shallow — they cover what’s easily Googlable and present generic overviews. This skill forces Claude to go far beyond that by executing a structured 5-layer research methodology with analytical frameworks, synthesis techniques, and built-in quality gates.

The 5 Research Layers:

1. Surface Scan — The basics (table stakes, done fast)
2. Structure & Landscape — How the space is organized, with decision logic and tradeoffs
3. Depth Dive — Hidden insights, failure stories, implementation gotchas, practitioner reality (where most research stops, this skill keeps going)
4. Adjacent Opportunities — Things you didn’t ask about but should know exist
5. Horizon Pointers — Brief flags for further exploration

Includes:

• 12 analytical frameworks with full application instructions (PESTEL, Porter’s Five Forces, JTBD, Competitive Positioning Map, SWOT, and more)
• Synthesis engine with triangulation, pattern recognition, confidence calibration, and a mandatory red-team protocol
• 6 topic-specific playbooks (Market Entry, Product Research, Competitive Analysis, Regulatory, Problem-Solving, Service Expansion)
• Good vs. bad output examples with commentary
• Dynamic source selection with weighting by topic type

Installation

Claude Desktop / Claude.ai

1. Download the deep-research.skill file from Releases
2. Drag and drop the .skill file into your Claude conversation
3. The skill is now active for that project

Manual Installation

1. Clone this repo or download the deep-research/ folder
2. Place the entire deep-research/ folder in your Claude skills directory
3. The skill triggers automatically on research-intent prompts

File Structure

deep-research/
├── SKILL.md                          # Core engine (mode detection, execution flow, output template)
└── references/
    ├── frameworks.md                 # 12 analytical frameworks with application guides
    ├── synthesis-engine.md           # Synthesis, red-team, confidence calibration
    ├── playbooks.md                  # 6 step-by-step research playbooks
    ├── examples.md                   # Good vs bad output comparisons
    └── source-selection.md           # Source categories, weighting, red flags

Usage

The skill triggers automatically when you use research-intent language:

"Research the utility discount market for disability placard holders"
"Dig into how referral features work in patient portals"
"What does the competitive landscape look like for telehealth GLP-1 services?"
"Explore whether we should expand into FMLA certifications"
"Help me understand the regulatory requirements for medical window tint exemptions"

Two Modes

Guided Mode — If your prompt is broad or ambiguous, Claude asks 3-5 targeted questions before executing.

Autonomous Mode — If your prompt is specific and detailed, Claude executes immediately.

You can also force a mode:

• “Ask me questions first, then research X” → Guided
• “Just go — research X” → Autonomous

What Makes This Different

Customization

The skill is designed to be modified. Common customizations:

• Add industry-specific playbooks — Copy the playbook template in references/playbooks.md and add steps specific to your industry
• Adjust framework selection — Modify the framework selection table in SKILL.md to match your most common research scenarios
• Add source categories — Add industry-specific sources to references/source-selection.md
• Tune depth — Adjust quality gates in each layer to match your depth requirements

License

MIT License — see LICENSE for details.