harness-bootstrap

Generates CLAUDE.md maps, docs/ skeletons, and .gitignore rules to initialize project harness structures.

Updated Jun 18, 2026
One-click install
npx skills add https://github.com/zhengguangli/harness-engineering-kit --skill harness-bootstrap-zhengguangli
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: harness-bootstrap
Source: https://github.com/zhengguangli/harness-engineering-kit/tree/main/skills/harness-bootstrap
Command: npx skills add https://github.com/zhengguangli/harness-engineering-kit --skill harness-bootstrap-zhengguangli

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? New projects lack the structured knowledge layer (CLAUDE.md routing map, docs/ directory, .gitignore rules) that AI agents need to navigate a codebase reliably, and hand-creating this scaffolding is repetitive and error-prone. ## Core Features & Use Cases - Map-Style CLAUDE.md Generation: Creates a concise routing table with hard constraints and workflow tips, keeping the file under 100 lines instead of an encyclopedia. - Minimum Viable docs/ Skeleton: Scaffolds ARCHITECTURE.md, QUALITY_SCORE.md, design-docs index, and exec-plans directories with "last updated" dates. - Project-Type Tailoring: Adjusts initialization scope for single-file scripts, small/medium/large applications, and monorepos, with per-tech-stack templates for Node.js, Python, Go, Rust, Java, PHP, Ruby, C#, Dart, and Elixir. - Use Case: You clone a FastAPI microservice with no agent-facing documentation. Run this Skill to analyze the project, generate a CLAUDE.md routing table pointing to a fresh docs/ skeleton, and append Python .gitignore rules, then verify with a 6-item checklist. ## Quick Start Initialize the harness structure for this project by generating a CLAUDE.md map, docs/ skeleton, and updated .gitignore rules.

Frequently Asked Questions about harness-bootstrap

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

FAQPage Schema
How do I initialize a CLAUDE.md file for a new project?▼

Run the harness initialization flow, which first analyzes your tech stack and structure, then generates a map-style CLAUDE.md containing a one-line description, hard constraints, a routing table to docs/ files, and workflow tips kept under 100 lines.

What files should a minimal docs/ directory contain for AI agents?▼

The minimum viable set is ARCHITECTURE.md for domain decomposition and dependency direction, QUALITY_SCORE.md for per-module quality tracking, a design-docs index, and exec-plans/active and exec-plans/completed directories. Every file needs a last-updated date at the bottom.

Does harness initialization work for monorepos with multiple tech stacks?▼

Yes. Monorepos get a unified docs/ skeleton at the root with shared ARCHITECTURE.md covering cross-package boundaries, per-sub-package routing entries in CLAUDE.md, and merged .gitignore rules for each detected stack such as Node.js and Go.

What happens if my project already has a CLAUDE.md or docs/ directory?▼

Existing content is read first, then you choose between overwriting or incremental updates. Stale structures are repaired item-by-item against a six-item checklist rather than reinitialized, and encyclopedia-style CLAUDE.md files are delegated to restructuring instead of being overwritten.

When should I not run full harness initialization?▼

Skip it when the project already has a complete harness structure, when you only want to understand the methodology, when the project is an extremely small script, or when you only need to restructure existing CLAUDE.md/docs content, which belongs to the repo-map skill.