Verified Research
Documentation Discovery - Use docs-seeker
For ALL documentation research tasks, ALWAYS use the docs-seeker skill.
docs-seeker provides comprehensive strategies:
- •llms.txt-first approach with context7.com prioritization
- •GitHub repository analysis via Repomix
- •Parallel exploration with multiple agents
- •Multi-source documentation gathering
Core Research Principles
Never present findings without examining actual source content.
Steps:
- •Fetch the actual source (WebFetch or extract tools)
- •Read the complete relevant sections
- •Verify claims match what source actually says
- •Quote specific passages when making claims
Documentation workflow:
- •Launch docs-seeker skill for target library/framework
- •Follow docs-seeker's multi-phase discovery process
- •Use context7.com llms.txt URLs when available
- •Verify all findings against source content
- •Present consolidated, verified information
Decision Framework
Use docs-seeker when researching:
- •✅ Any library, framework, or technology
- •✅ API documentation and references
- •✅ Installation guides and setup instructions
- •✅ GitHub repositories with documentation
- •✅ Technical specifications and best practices
- •✅ When user provides library name but no specific URLs
Use general research when:
- •✅ User provides specific URLs to study
- •✅ Studying source code directly
- •✅ Verifying claims from search snippets
- •✅ Research topics without clear documentation focus
- •✅ Exploring non-documentation sources
Common Pitfalls
❌ Presenting search snippets as facts ❌ Trusting summaries without checking sources ❌ Citing sources you haven't read
When Uncertain
If you can't verify (paywall, 404, contradictions): Say so explicitly. Don't present unverified info as fact.
References
For detailed patterns and examples:
- •references/verification-patterns.md
- •references/repo-cloning-pattern.md - Clone repos via subagent for source-level research