harness-design CLAUDE.md

harness-design CLAUDE.md is an instructions file for coding agents from zanwei/harness-design. It costs 696 tokens per session, scanned A, original, MIT.

A verification workflow for checking whether generated HTML matches a design description and a reference website. It uses several checks, including browser screenshots and technical inspections, to find differences and guide corrections.

In plain words
What is it for?
Use it after a design specification and HTML page have been created, to inspect resources, compare the result with the original site, score the match, and identify fixes.
Why use it?
It helps catch visual and implementation differences that ordinary code review may miss when rebuilding a website from a reference.

Instructions file

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add instructions/zanwei/harness-design/claude-md
Clone the repo
git clone --depth 1 https://github.com/zanwei/harness-design

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for harness-design CLAUDE.md

README.md
[![agentmods](https://agentmods.dev/badge/instructions/zanwei/harness-design/claude-md.svg)](https://agentmods.dev/instructions/zanwei/harness-design/claude-md)
Your own site
<a href="https://agentmods.dev/instructions/zanwei/harness-design/claude-md"><img src="https://agentmods.dev/badge/instructions/zanwei/harness-design/claude-md.svg" alt="Measured on agentmods" height="20"></a>
Per session 696 This file is loaded in full into every session.
When invoked 696 The same file — it is already loaded in full.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00696 $0.00696
Opus 5 $0.00348 $0.00348
Sonnet 5 $0.00139 $0.00139
Haiku 4.5 $0.00070 $0.00070

Measured 5d ago against content hash d8a60964f5e9, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

harness-design CLAUDE.md scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 5d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

CLAUDE.md · 53 lines

How it starts

The opening of the file, as written. The whole thing — 53 lines — stays where its author put it; the contents beside it link to each section on GitHub.

CLAUDE.md

Claude Code marketplace plugin for design-dna downstream QA. Bundles a single skill that scores generated HTML against both a Design DNA JSON spec and the original reference website using browser-based visual verification.

Root README files follow design-dna: README.md (English), README.zh-CN.md, README.ja.md, README.ko.md, README.es.md, README.zh-TW.md, with the same centered title and cross-linked language bar.

Architecture

Path Role
.claude-plugin/marketplace.json Plugin metadata, version, and skill paths
skills/harness-design/ SKILL.md + references/ (rubric, verification checklist, browser procedures)

The skill uses browser MCP tools (navigate, screenshot, network_requests, console_messages, profiling) and optionally the web-shader-extractor Playwright scripts for resource collection.

Workflow

  1. design-dna Phase 2 → DNA JSON (from reference URL)
  2. design-dna Phase 3 → HTML output
  3. harness-design → 5-pass verification (resource audit + mechanical + visual comparison + perceptual + effects), weighted score, correction loop

The harness does not edit the DNA JSON; it only adjusts generated markup/CSS/JS to match the spec and reference.

5-Pass Verification

Pass Name Tools Used
0 Resource Audit fetch-rendered-dom.mjs (Playwright) or browser_network_requests + browser_console_messages
1 Mechanical Checks browser_snapshot + getComputedStyle injection + DNA JSON matching
2 Visual Comparison browser_take_screenshot + browser_scroll (section-by-section)
3 Perceptual Review Screenshot-informed qualitative assessment
4 Effects Audit Frame sampling + browser_profile_start/stop + auxiliary skills

Auxiliary Skills

Effect Type Skill Usage
WebGL / Shader / Canvas web-shader-extractor fetch-rendered-dom.mjs for resource collection, scan-bundle.sh for framework ID, shader param extraction
ScrollTrigger / GSAP gsap-scrolltrigger Scroll trigger position identification, pin/scrub verification
Three.js / 3D 3d-web-experience 3D engine analysis, rendering verification

Read the full file on GitHub · 53 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 5d ago First seen · 53 lines · 696 tokens per session scan A d8a60964f5e9

Subscribe to this mod's changes

harness-design CLAUDE.md is an instructions file published in the GitHub repository zanwei/harness-design (4 stars, last pushed 4mo ago), licensed MIT. It adds 696 tokens to every session, about $0.0035 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.

Related

Other instructions, from other repositories