| name | skill-researcher |
| description | Research concepts, technologies, and best practices from the internet to create or refine Claude Code skills with up-to-date, expert-level knowledge. Use when: (1) Creating a new skill and needing to research the subject first, (2) Refining an existing skill with current best practices, (3) User asks to "research [topic] for a skill", (4) User wants a skill informed by latest documentation and community patterns, (5) Enriching a skill with knowledge beyond Claude's training data.
|
Skill Researcher
Research topics from the web and synthesize findings into well-structured skill content.
Workflow
- Clarify scope - Identify the subject, target skill, and research goals
- Research - Search the web for authoritative, current knowledge
- Synthesize - Extract actionable patterns, filter for skill relevance
- Produce - Create or update the skill with researched knowledge
Step 1: Clarify Scope
Determine before researching:
- Subject: What topic/technology/concept to research
- Target: Creating a new skill or refining an existing one?
- Goals: What specific knowledge gaps to fill (best practices, patterns, anti-patterns, configuration, API details)
If refining an existing skill, read it first to identify gaps.
Step 2: Research
Use WebSearch and WebFetch to gather knowledge. See references/research-strategies.md for query patterns and source evaluation criteria.
Research phases
Phase 1 - Official sources: Search for official documentation, guides, and changelogs. Fetch and extract key patterns.
Phase 2 - Community wisdom: Search for blog posts, conference talks, and real-world usage patterns. Focus on battle-tested advice.
Phase 3 - Anti-patterns: Search for common mistakes, pitfalls, and migration issues. These are high-value for skills.
Research guidelines
- Run 3-6 targeted searches per phase (not more - diminishing returns)
- Prefer recent content
- Cross-reference findings across multiple sources
- Capture concrete code examples, not just descriptions
- Note version-specific details (breaking changes, deprecations)
Step 3: Synthesize
Filter research through these criteria:
Include knowledge that is:
- Non-obvious to Claude (procedural, version-specific, configuration)
- Actionable (changes how code is written)
- Reusable (applies across multiple tasks)
Exclude knowledge that is:
- General programming advice Claude already knows
- Overly theoretical without practical application
- Outdated or contradicted by multiple sources
Organize findings into:
- Core patterns (go in SKILL.md body)
- Detailed references (go in references/ files)
- Code templates (go in assets/ if substantial)
Step 4: Produce
New skill
Follow the skill-creator workflow. Use research findings as the primary content source. Structure using progressive disclosure - core knowledge in SKILL.md, details in references/.
Refining existing skill
- Read the current skill files
- Identify where research findings improve or complement existing content
- Update SKILL.md and reference files with new knowledge
- Remove outdated information replaced by research findings
Quality checks
- Every pattern includes a concise code example
- Anti-patterns explain why to avoid (not just what)
- Version/compatibility notes are explicit
- Content follows skill-creator guidelines (concise, imperative form, progressive disclosure)
