compress-skill

To compress markdown documentation without losing semantics, apply deterministic cleanup to remove blank lines and meta prose, then use optional LLM-based rewrites to shrink text while preserving meaning.

The best way to shrink a SKILL.md file is to run deterministic cleanup followed by an optional LLM rewrite, which reduces token usage while keeping the document concise for efficient context loading.

You can reduce markdown verbosity while preserving frontmatter and code blocks by using invariants checks that ensure anchors, headings, and fenced code blocks remain completely intact during the compression process.

Yes, you can test markdown compression changes safely by running a dry-run, which previews the deterministic cleanup and LLM rewrite results without modifying the original file until you explicitly use the apply command.

LLM-based markdown compression maintains semantic integrity by engaging only when necessary to meet target compression limits, ensuring the original meaning is preserved while reducing text length.

Deterministic cleanup is limited to collapsing blank lines, trimming trailing whitespace, and removing meta prose, meaning it cannot rewrite sentences or reduce text length semantically without the optional LLM component.