markdown-compressor

Recursively find and compress all matching markdown files in a repository — auto-approve mode with batched processing

Compress a markdown file to reduce token usage — lossless (structural) or lossy (semantic) modes with section-by-section review

Use this agent to review a compressed markdown section against its original. This agent catches information loss, over-aggressive compression, and broken references. It is spawned by the /compress command after the section-compressor — not invoked directly by users. <example> Context: The /compress command just received a compressed section from section-compressor and needs quality review. user: "Compress my CLAUDE.md file" assistant: "I'll dispatch the compression-reviewer agent to verify no critical information was lost." <commentary> After section-compressor returns a compressed version, the compression-reviewer compares original and compressed text to catch information loss before presenting to the user. </commentary> </example> <example> Context: Lossy compression of an agent instruction file — reviewer checks the aggressive compression. user: "/compress agents/code-reviewer.md" assistant: "The compression-reviewer agent will verify the compressed version preserves all behavioral rules and specifics." <commentary> The reviewer is especially critical in lossy mode where semantic changes are expected. It verifies that essential information survived compression. </commentary> </example>

Use this agent to compress a single markdown section during the /compress workflow. This agent receives one section at a time and returns the compressed version. It is spawned by the /compress command — not invoked directly by users. <example> Context: The /compress command is processing a CLAUDE.md file section-by-section in lossy mode. user: "Compress my CLAUDE.md file" assistant: "I'll dispatch the section-compressor agent to compress this section." <commentary> The /compress command splits the file into sections and dispatches section-compressor for each one. The agent applies compression techniques from the markdown-compression skill and returns the compressed text. </commentary> </example> <example> Context: The /compress command is processing an agent instruction file in lossless mode. user: "/compress agents/code-reviewer.md --lossless" assistant: "I'll use the section-compressor agent to apply lossless transformations to this section." <commentary> In lossless mode, the section-compressor applies only structural transformations. In lossy mode, it applies semantic compression. The mode is specified in the agent's task prompt. </commentary> </example>

This skill should be used when the user asks to 'compress markdown', 'shrink this file', 'optimize tokens', 'reduce file size', 'compress instructions', 'make this more concise', 'minimize this prompt', 'compress CLAUDE.md', 'compress ARCHITECTURE.md', 'optimize agent instructions', or wants to reduce token usage in LLM-facing markdown files. Covers lossless structural optimization and lossy semantic compression with section-by-section analysis.

1 hook across 1 event