content-retrieval

Section-aware retrieval uses markdown headings and horizontal rules to extract targeted content. Specify the section by heading name, and the Skill returns only that portion, minimizing token usage and read time for large files.

Depth levels L0–L5 offer graduated retrieval operations: Exists checks file presence, Count tallies lines, Lookup finds specific data, Overview scans structure, Section isolates content by heading, and Full reads the entire file. Choose the minimal depth required to answer your question and avoid unnecessary tokens.

Yes. Token-efficient retrieval supports offset and limit parameters to extract a specific range of lines or content blocks. Combined with depth-level selection, this approach fetches only the data you need while preserving context options for surrounding lines.

The Skill detects section boundaries using markdown heading syntax and horizontal rules, then maps content to those sections. Query by heading name to fetch that section directly without scanning the entire file structure.

Depth-aware retrieval is format-agnostic and applies across file reads. Markdown-based section detection requires markdown heading syntax; grep and line-based operations work on any text file, making the technique adaptable to logs, configs, and structured documents.

Use Lookup when searching for a specific value or pattern without needing context. It returns only matching lines, drastically reducing tokens compared to Full read. Reserve Full read for comprehensive analysis when overview or targeted section retrieval won't suffice.