Unified Search Skill
Search across code AND documents, with automatic cross-reference detection.
Quick Start
# 1. Index a project (code + docs) /wicked-search:index /path/to/project # 2. Search everything /wicked-search:search "authentication" # 3. Find where a symbol is documented /wicked-search:refs AuthService # 4. Find code that implements a doc section /wicked-search:impl "Security Requirements"
Commands Reference
| Command | Purpose |
|---|---|
/wicked-search:index <path> | Build index for code + docs |
/wicked-search:search <query> | Search everything |
/wicked-search:code <query> | Search code only |
/wicked-search:docs <query> | Search docs only |
/wicked-search:refs <symbol> | Find references + documentation |
/wicked-search:impl <section> | Find implementing code |
/wicked-search:scout <pattern> | Quick pattern recon (no index needed) |
/wicked-search:stats | Show index statistics |
When to Use Each Command
User asks about code
Use /wicked-search:code for:
- •"Find the UserService class"
- •"Where is authenticate defined?"
- •"Search for functions that handle login"
User asks about documents
Use /wicked-search:docs for:
- •"What does the requirements doc say about X?"
- •"Find the API spec section on authentication"
- •"Search the design docs for rate limiting"
User asks about relationships
Use /wicked-search:refs for:
- •"Where is this function documented?"
- •"What calls this class?"
- •"What does this module import?"
Use /wicked-search:impl for:
- •"What code implements the auth spec?"
- •"Find the code for this requirement"
User asks general questions
Use /wicked-search:search for:
- •Broad searches across everything
- •When unsure if it's code or docs
- •"Find anything related to authentication"
Cross-Reference Detection
The indexer automatically detects when documents mention code symbols:
- •CamelCase:
AuthService,UserRepository - •snake_case:
authenticate_user,get_user_by_id - •Backtick quoted:
`my_function` - •Function calls:
process_data()
These create "documents" edges in the graph, linking doc sections to code.
Reading Extracted Documents
After indexing, extracted document text is cached at:
~/.something-wicked/search/extracted/<filename>.txt
Read these files to get full document content for context.
Tips
- •Index first: Always run
/wicked-search:indexbefore searching - •Be specific: More specific queries = better results
- •Check cross-refs: Use
/refsto discover code-doc relationships - •Read cached docs: For full context, read the extracted .txt files
Supported File Types
Code: Python, JavaScript, TypeScript, Java, Go, Rust, C/C++, Ruby, and 20+ more
Documents: PDF, Word (.docx), Excel (.xlsx), PowerPoint (.pptx), Markdown, HTML, plain text