write-claude-md

Writes or slims down a project's CLAUDE.md through repo mapping and user interview.

Updated Aug 7, 2026
One-click install
npx skills add https://github.com/sagittaras/agentic-workflow --skill write-claude-md-sagittaras
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: write-claude-md
Source: https://github.com/sagittaras/agentic-workflow/tree/main/skills/write-claude-md
Command: npx skills add https://github.com/sagittaras/agentic-workflow --skill write-claude-md-sagittaras

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? CLAUDE.md files written by hand or by the default /init command tend to bloat with redundant command listings, folder inventories, and generic advice, wasting context budget in every session. This Skill produces a concise, intent-focused CLAUDE.md by mapping the repository, interviewing the user about project purpose, and enforcing a strict line budget. ## Core Features & Use Cases - Repository mapping: Scans structure, tech stack manifests, and documentation locations without reading code, producing a signpost-style overview rather than an inventory. - Intent interview: Asks the user open questions about project purpose, domain, and architectural principles — the content that cannot be derived from code. - Outline approval gate: Presents a section-by-section outline with explicit omissions for user approval before writing anything. - Size enforcement: Targets ~80 lines with a hard 200-line ceiling, cutting sections rather than diluting them, and escalating when the budget cannot be met. - Use Case: A project's CLAUDE.md has grown to 400 lines of stale build commands and folder listings. Invoke the Skill to slim it down: it reads the existing file fully, preserves valid decisions, and rewrites it as a lean signpost document. ## Quick Start Ask the assistant to write or slim down this project's CLAUDE.md, optionally noting what the file should emphasize.

Frequently Asked Questions about write-claude-md

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I write a good CLAUDE.md for my project?

Use this Skill instead of the built-in /init command. It maps your repository structure and tech stack, interviews you about the project's purpose and principles, presents an outline for approval, and only then writes the file targeting roughly 80 lines.

How to slim down an overly long CLAUDE.md file?

The Skill reads the existing file completely, treats it as input rather than waste, and preserves valid decisions while cutting bloat. It reduces size by deleting sections replaceable with documentation links, never by diluting content into vaguer wording.

What should not go into a CLAUDE.md file?

Exclude build and test commands readable from manifests, full folder inventories, explanations of well-known technologies, copied documentation content, generic code-quality advice, formatting rules, dependency lists, and dated roadmaps. The bundled template reference lists each exclusion with its rationale.

Does the Skill handle both CLAUDE.md and .claude/CLAUDE.md locations?

Yes. It detects both root CLAUDE.md and .claude/CLAUDE.md, writes where the file already exists, and if both exist proposes a merge plan. Deleting the redundant file requires the user, since the Skill has no deletion tool.

When should I not use this Skill for project documentation?

Do not use it for moving content into path-scoped rules (use write-rule), creating skills or agents, writing architecture decision records, or art direction documents. It is scoped specifically to the project-level CLAUDE.md instruction file.