harness:writing-skills

Write and validate process documentation skills using TDD and YAML frontmatter.

3|1|Updated Apr 5, 2026
One-click install
npx skills add https://github.com/huangbaixun/harness-engineering --skill harness-writing-skills
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: harness:writing-skills
Source: https://github.com/huangbaixun/harness-engineering/tree/main/skills/writing-skills
Command: npx skills add https://github.com/huangbaixun/harness-engineering --skill harness-writing-skills

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

It prevents unreliable documentation by standardizing how agents author and validate new skills so the right behavior is consistently followed under pressure.

Core Features & Use Cases

  • TDD-style skill authoring for documentation: Create pressure-tested rules for when to use a skill, then verify agents comply instead of rationalizing.
  • Metadata that improves discovery (CSO): Produce YAML frontmatter that makes Claude reliably find the right skill based on triggering conditions, symptoms, and keywords.
  • Vendoring discipline and eval-gating: Enforces the harness workflow so skill changes are not committed without corresponding eval coverage where required.

Quick Start

Use harness:writing-skills to design and verify a new or updated skill by specifying its triggering conditions and then running baseline-vs-with-skill eval scenarios before deployment.

Frequently Asked Questions about harness:writing-skills

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

FAQPage Schema
How do I write process documentation that AI agents reliably follow?

To write process documentation agents reliably follow, use a TDD-based authoring method that defines triggering conditions, enforces iron rules, and validates agent compliance using eval scenarios before deployment.

What is the best way to make Claude Code discover my custom skills?

The best way to make Claude Code discover custom skills is adding CSO-friendly YAML frontmatter with harness namespacing, triggering conditions, symptoms, and keywords to your instruction documents.

How do I validate new skills for Claude Code before deploying them?

You validate new skills by running baseline-vs-with-skill eval scenarios and applying structure requirements to verify agents comply with the rules instead of rationalizing behavior.

Why does my YAML frontmatter for agent skills fail discovery checks?

YAML frontmatter fails discovery checks when description text avoids workflow summaries, lacks the Use when prefix, or misses harness namespacing required for CSO discovery metadata compliance.

Do I need eval coverage for every skill authoring change?

You need eval coverage for skill authoring changes when required by the harness eval and checklist workflow, enforcing vendoring discipline so commits include corresponding eval validation.