📄 File Operations

👤 Who it's for: Developers / data wranglers
⏱️ Read time: ~5 minutes
💡 In one line: AI reads, writes, searches, and batch-edits files directly — search big files first, confirm before edit, all safe and controllable.

YingClaw provides full read, write, edit, and search capabilities for files, covering everyday development and operations scenarios.

Core Tools

file_read — Read File

Read any text file, with support for segmented reading and automatic PDF extraction.

Parameter	Type	Description
path	string	File path (relative to workspace)
offset	number	Starting line number (1-based), optional
limit	number	Maximum lines to read, optional

# Read entire file
file_read path="config.toml"

# Read in segments (start at line 100, read 50 lines)
file_read path="app.log" offset=100 limit=50

file_write — Write/Create File

Create a new file or overwrite an existing one. Parent directories are created automatically.

Parameter	Type	Description
path	string	File path
content	string	Full content to write

# Create README.md
file_write path="docs/guide.md" content="# User Guide\n\n..."

file_edit — Precise Replacement

Find a unique matching string in a file and replace it with new content.

Parameter	Type	Description
path	string	Path of file to edit
old_string	string	Original string to replace (must match exactly once)
new_string	string	Replacement string

Note: old_string must appear exactly once in the file, otherwise the edit fails. Include enough surrounding context to ensure uniqueness.

glob_search — File Search

Match file paths by glob pattern.

# Search all Rust source files
glob_search pattern="**/*.rs"

# Search module files under src
glob_search pattern="src/**/mod.rs"

content_search — Content Search

Search file contents using regex, powered by ripgrep.

# Search for 'useState' in all .ts files
content_search pattern="useState" include="*.ts"

# Search and show match counts
content_search pattern="TODO" output_mode="count"

Use Cases

Scenario	Recommended Tool
Code review	file_read + content_search
Config changes	file_read → file_edit
Log analysis	content_search + file_read (segmented)
Document generation	file_write
Data extraction	file_read + glob_search
Batch refactoring	glob_search → file_edit

Best Practices

Large File Strategy

File Size	Strategy
< 50KB	Use file_read to read entire file
50KB ~ 500KB	Use offset + limit for segmented reading
> 500KB	Never read in full, use content_search for key content only

Search First, Then Deep Read

For large projects: use glob_search to locate target files, then content_search to find key lines, and finally file_read for precise reading.

Read Before Editing

Always file_read to confirm current content before editing. Construct a precise old_string to avoid accidental changes.

Special Format Support

• PDF files: file_read can auto-extract PDF text content (requires rag-pdf feature)
• Code files: Auto-detect language and preserve syntax highlighting context
• Config files: Structured reading for JSON / YAML / TOML formats

Next Steps

Master file operations, then learn about 🧠 Memory System to let YingClaw remember your preferences.