Split large documents into atomic shards for efficient context management.
npx skills add https://github.com/alessioroberto82/claude-plugin-bmad --skill shard
Please help me install this Agent Skill. Skill: shard Source: https://github.com/alessioroberto82/claude-plugin-bmad/tree/main/plugin/skills/shard Command: npx skills add https://github.com/alessioroberto82/claude-plugin-bmad --skill shard
This Skill tackles the issue of large documents consuming excessive tokens, making AI interactions slow and expensive, by breaking them down into smaller, manageable pieces.
Use the shard skill to break down the document located at ~/.claude/circle/projects/myproject/output/prioritize/PRD-final.md into smaller shards.
High-intent search queries and answers about installing and using this skill.
Context sharding splits large documents like architecture specifications into atomic files. This technique manages context by generating distinct shards for requirements, architecture decisions, and user stories following a defined naming convention.
Context sharding splits large documents like architecture specifications into atomic files. This technique manages context by generating distinct shards for requirements, architecture decisions, and user stories following a defined naming convention.
To extract a user story without loading the entire PRD, run the shard skill against your document path. It automatically detects input documents from specified project output directories and breaks them down into manageable pieces.
Yes, automatic document detection is supported from specified project output directories. You simply point the sharding process to a path like your project's output folder, and it generates distinct shards automatically.
The best way to save tokens with large architecture docs is context sharding, which divides them into atomic files. This enables AI roles to load only relevant sections, achieving up to 90% token reduction.
Sharding large documents does not randomize outputs; it generates distinct shards for requirements, architecture decisions, and user stories by following a defined naming convention, ensuring organized and traceable file management.
ncptuanSplit large Markdown documents into section-based files by level 2 headings.
1ambdaLocate functions, classes, and patterns in codebases using MCP tools.
DbillionaerBundle user prompts with selected project files for AI code generation.
duy-tungOrganize project files into structured directories by type and naming conventions.
