humanjs CLAUDE.md

Project instructions for HumanJS, a browser-automation project designed for AI agents, quality checks, and recording demonstrations. They also define the project's scope and work that must be refused.

In plain words
What is it for?
Use them when making decisions or writing material for the HumanJS monorepo—a repository that contains multiple related packages or projects—to preserve its positioning, status, and non-goals.
Why use it?
They keep code, documentation, and examples aligned with the project's intended audience and prevent work on excluded areas such as CAPTCHA bypass or fingerprint masking.

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/totigm/humanjs/claude-md
Clone the repo
git clone --depth 1 https://github.com/totigm/humanjs
Per session 3,451 This file is loaded in full into every session.
When invoked 3,451 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 $0.03451 $0.03451
Opus 5 $0.01725 $0.01725
Sonnet 5 $0.00690 $0.00690
Haiku 4.5 $0.00345 $0.00345

Measured 3d ago against content hash c90a0e65fe3d, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

humanjs 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 3d 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 · 250 lines

How it starts

The opening of the file, as written. The whole thing — 250 lines — stays where its author put it; the contents beside it link to each section on GitHub.

HumanJS

Humanized browser automation for AI agents, QA tests, and demos. Playwright-first.

This file is the brief loaded into every Claude session in this repo. Keep it lean. Architectural reasoning lives in docs/DESIGN.md (public). Strategic and personal notes live in docs/PROJECT_BRIEF.md (local-only, gitignored).

Project status

Greenfield — no code yet. We're scaffolding the v1 monorepo from scratch.

Positioning (immutable)

HumanJS humanizes browser automation for AI agents, QA tests where real-pace timing exposes bugs, and demo/tutorial recordings.

When writing copy, docs, examples, or comments: never frame the project around scraping, captcha bypass, or "undetectable" automation. The audience is AI agent builders, QA engineers, and demo/tutorial creators.

Non-goals — refuse to implement

If asked, decline and link to this section.

  • Captcha solvers / captcha bypass
  • Fingerprint masking or browser fingerprint spoofing
  • Proxy rotation or IP management
  • TLS / network-level stealth
  • "Undetectable" claims anywhere in any artifact
  • Time-of-day variation (gimmick, no real value)
  • AI-generated mouse trajectories (worse than Bezier, vastly more expensive)

Naming

  • Brand: HumanJS
  • Primary npm scope: @humanjs/*
  • Fallback scope: @totigm/humanjs (used only if @humanjs org isn't claimable on first publish)
  • Domain: humanjs.dev (planned)

Verified available on npm at project start: @humanjs/core, @humanjs/playwright, @humanjs/recorder, @humanjs/mcp, @humanjs/generator, @humanjs/skill, @humanjs/recipes. Unscoped humanjs is owned by a dormant 2022 package — irrelevant scope, doesn't block us.

Architecture

  • Monorepo (pnpm workspaces, optionally Turborepo). Each package independently versioned.
  • Playwright-first. Puppeteer adapter is roadmap, not v1.
  • Plugin system from day one. Even if v1 ships no plugins, the architecture supports extending personalities, primitives, and behaviors via plugins. Non-negotiable architectural decision.
  • TypeScript strict. ESM-first with CJS dual-publish where needed.
  • MIT licensed.

Read the full file on GitHub · 250 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. 3d ago First seen · 250 lines · 3,451 tokens per session scan A c90a0e65fe3d

Subscribe to this mod's changes

humanjs CLAUDE.md is an instructions file published in the GitHub repository totigm/humanjs (11 stars, last pushed 3d ago), licensed MIT. It adds 3,451 tokens to every session, about $0.0173 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 2026-08-30.