long-document-handler

Large documents exceeding 20,000 tokens can cause incomplete analysis and truncated responses. This skill automatically detects oversized content and pauses processing, notifying you of the document's size and offering alternatives like targeted extraction, summarization, or sharding to keep Claude within safe limits.

When a document surpasses token limits, you can choose between multiple strategies: extract specific sections you need, generate a high-level summary, split the document into manageable shards, or proceed with full reading if context permits. Each option is presented with explicit thresholds (80,000+ characters, 15,000+ words, 2,000+ lines, 20,000+ tokens).

Yes, this skill is designed for lengthy structured documents like architecture documentation, product requirement documents, and extended markdown files. It detects when these files approach context limits and offers processing modes tailored to complex, multi-section content.

The skill estimates tokens using a roughly 4-character-per-token ratio and flags documents when they reach 20,000 tokens or equivalent thresholds (80,000+ characters, 15,000+ words, 2,000+ lines), giving you an upfront warning before processing stalls.

This skill supports guided strategies for writing large documents in phases or split designs, but the actual composition happens in stages. It helps structure complex content and integrates with shard and BMAD workflows, though it enforces stop-on-threshold behavior rather than enabling unlimited writing.

Sharding splits your document into smaller, independently processable chunks, preserving detail and structure across the entire content. Summarization provides high-level overviews when detail isn't needed. Choose sharding for comprehensive analysis of large files and summarization when you need quick insights without full content.