humanize-automation
Human-like mouse, keyboard and scroll behavior for behavioral bot bypass.
Install / Use
npx skills add uphiago/recon-skills --skill humanize-automationInstalls into whichever agent you are using.
SKILL.md
Installable skill definition
Quality Score
Category
AutomationSupported Platforms
Our assessment of humanize-automation
humanize-automation scores 89/100 on our quality scale, 1383rd of 2,780 Automation skills we index (top 50%).
Its SKILL.md is 8.8 KB long, well organised into 26 sections with 6 code examples: a thorough specification that gives an agent plenty to work with.
With 1,280 GitHub stars, it is one of the more widely adopted skills in the catalogue.
Maintenance, license and trust
- The repository was last updated about 3 months ago, so humanize-automation is actively maintained.
- Our last check on 2026-10-31 found the source still online.
- It is released under the MIT license, a permissive license that allows use, modification and commercial use with attribution.
- Its trust signals score 100/100, with no cautions. These come from repository metadata, not a code audit — read the skill file before letting an agent act on it.
Safety scan
No issues foundOur scan of the whole file found no instruction hijacking, hidden characters, credential access, data exfiltration or destructive commands. An AI review of the same text found nothing harmful.
AI review by kimi-k2.7-code on 2026-10-01. Automated pattern scan on 2026-10-01. It catches known dangerous patterns, not every risk — read a skill before letting an agent act on it.
humanize-automation compared with similar skills
All 4 of these similar skills score higher than humanize-automation; compare them before choosing.
| Skill | Score | Stars | Updated | Format |
|---|---|---|---|---|
| humanize-automation (this skill)by uphiago | 89 | 1.3k | 3mo ago | SKILL.md |
| Agent-Reachby Panniantong | 100 | 112.3k | 9d ago | CLAUDE.md |
| rufloby ruvnet | 100 | 81.9k | today | CLAUDE.md |
| headroomby headroomlabs-ai | 100 | 79.8k | today | CLAUDE.md |
| Claude-Code-Game-Studiosby Donchitos | 100 | 39.7k | 7d ago | CLAUDE.md |
Frequently asked questions
- How do I install humanize-automation?
- Run
npx skills add uphiago/recon-skills --skill humanize-automation. The install tabs above show the steps for each supported agent. - Which AI agents does humanize-automation work with?
- It is written for Universal, as a SKILL.md file. Other agents that read the same format can often use it too.
- Is humanize-automation safe to use?
- Our scan of the whole file found no instruction hijacking, hidden characters, credential access, data exfiltration or destructive commands. An AI review of the same text found nothing harmful. It is MIT-licensed and scores 100/100 on trust signals. Skills are instructions an agent will follow, so read the file before installing it and do not approve commands you do not understand.
- Is humanize-automation still maintained?
- The repository was last updated about 3 months ago, so humanize-automation is actively maintained.
Skill content
View source on GitHubname: humanize-automation description: Human-like mouse, keyboard and scroll behavior for behavioral bot bypass. version: 1.1.0 revision_date: 2026-07-25 license: MIT platforms: [linux] compatibility: Requires python3 tags: [recon, humanize, behavioral, anti-bot, mouse, keyboard, automation] category: recon related_skills:
- stealth-browser-launch
- tls-fingerprint-impersonation
Humanize Automation
Replace instant programmatic interactions with human-like mouse movements, keyboard typing, and scroll patterns. Patches Playwright's API at the class level — page.click(), page.type(), page.fill(), and Locator methods are automatically replaced with Bézier-curved mouse paths, per-character typing with mistypes, and multi-phase scroll acceleration. One flag (humanize=True) enables all behavioral patches. No code changes required.
When to Use
- Target uses behavioral bot detection (mouse trajectory analysis, typing speed profiling).
- reCAPTCHA v3 scores are low (<0.3) despite correct browser fingerprint.
- Target times out or challenges after rapid form submissions.
- Need to simulate a real user browsing session for login or account creation.
- Target uses
requestAnimationFrame-based mouse movement tracking.
Prerequisites
terminalwith python3.cloakbrowserinstalled:pip install cloakbrowser.- Or standalone Playwright with custom patching: import
patch_pagefrom the humanize module.
Quick Start
from cloakbrowser import launch
browser = launch(humanize=True)
page = browser.new_page()
page.goto("https://target.com/login")
# All interactions are automatically humanized
page.locator("#email").fill("user@example.com") # per-character timing
page.locator("#password").fill("password123") # with thinking pauses
page.locator("button[type=submit]").click() # Bézier curve movement
browser.close()
Procedure
Phase 1 — Default Humanization
One flag enables all behavior patches:
browser = launch(
headless=False,
proxy="http://residential-proxy:port",
geoip=True,
humanize=True, # enables all behavioral patches
)
page = browser.new_page()
page.goto("https://target.com")
# All Playwright interactions are replaced with human-like equivalents
page.locator("input[name='search']").fill("restricted query")
page.locator("button[type='submit']").click()
Phase 2 — Careful Mode (Slower, More Deliberate)
For sites that profile interaction speed:
browser = launch(
humanize=True,
human_preset="careful", # slower typing, more idle pauses
)
Careful preset changes: typing 100±50ms (vs 70±40ms), click aim 80-180ms (vs 60-140ms), idle_between_actions enabled with 0.4-1.0s pauses.
Phase 3 — Custom Behavior Profile
Override individual parameters for site-specific tuning:
from cloakbrowser import HumanConfig
custom = HumanConfig(
typing_delay=150,
typing_delay_spread=60,
typing_pause_chance=0.15,
mistype_chance=0.03,
mouse_wobble_max=2.0,
click_aim_delay_input=(100, 200),
scroll_delta_base=(60, 100),
idle_between_actions=True,
idle_between_duration=(0.5, 1.5),
)
browser = launch(humanize=True, human_config=custom)
Phase 4 — Per-Call Overrides
Override behavior for individual interactions:
# Slow, careful typing for password field
page.locator("#password").fill("secret", human_config={"typing_delay": 200})
# Fast click (bypass aim delay)
page.locator("#submit").click(human_config={"click_aim_delay_input": (0, 0)})
Phase 5 — Standalone Humanization (Without CloakBrowser)
Patch an existing Playwright page:
from playwright.sync_api import sync_playwright
from cloakbrowser.human import patch_page, resolve_config
pw = sync_playwright().start()
browser = pw.chromium.launch()
page = browser.new_page()
# Apply humanization to an existing page
config = resolve_config("default")
patch_page(page, config)
# All interactions now humanized
page.locator("#email").fill("user@example.com")
Behavioral Parameters Reference
Mouse Movement
| Parameter | Default | Effect |
|---|---|---|
| mouse_steps_divisor | 8 | Controls smoothness (lower = smoother, more steps) |
| mouse_min_steps | 25 | Minimum steps per movement |
| mouse_max_steps | 80 | Maximum steps per movement |
| mouse_wobble_max | 1.5 px | Sideways sinusoidal wobble amplitude |
| mouse_overshoot_chance | 0.15 | Probability of overshooting target |
| mouse_overshoot_px | (3, 6) px | Overshoot distance |
Movement algorithm: Cubic Bézier curve with 2 random perpendicular control points → ease-in-out cubic easing → sinusoidal wobble → 15% overshoot chance with correction.
Keyboard Typing
| Parameter | Default | Effect |
|---|---|---|
| typing_delay | 70 ms | Average per-character delay |
| typing_delay_spread | 40 ms | Random spread around delay |
| typing_pause_chance | 0.10 | Thinking pause probability per character |
| typing_pause_range | (400, 1000) ms | Thinking pause duration |
| mistype_chance | 0.02 | Typo probability per character |
| mistype_delay_notice | (100, 300) ms | Time before noticing typo |
| mistype_delay_correct | (50, 150) ms | Time to correct typo |
| field_switch_delay | (800, 1500) ms | Delay when switching between form fields |
Typing algorithm: Per-character delay with random spread → 10% pause chance → 2% typo chance (nearby key → notice → backspace → retype) → Shift symbols via CDP Input.dispatchKeyEvent for isTrusted=true.
Scrolling
| Parameter | Default | Effect |
|---|---|---|
| scroll_delta_base | (80, 130) px | Scroll distance per step |
| scroll_accel_steps | (2, 3) | Acceleration phase steps |
| scroll_decel_steps | (2, 3) | Deceleration phase steps |
| scroll_overshoot_chance | 0.10 | Probability of overshooting |
| scroll_overshoot_px | (50, 150) px | Overshoot distance |
| scroll_target_zone | (0.20, 0.80) | Target zone in viewport |
| scroll_settle_delay | (300, 600) ms | Delay after reaching target |
Scroll algorithm: Accelerate (2-3 steps, 80-100px) → cruise (80-130px, burst into 20-40px wheel chunks at 8-20ms intervals) → decelerate (2-3 steps, 60-90px) → 10% overshoot chance with correction.
Click Behavior
| Parameter | Default | Effect |
|---|---|---|
| click_aim_delay_input | (60, 140) ms | Delay before clicking input elements |
| click_aim_delay_button | (80, 200) ms | Delay before clicking buttons |
| click_hold_input | (40, 100) ms | Hold duration on inputs |
| click_hold_button | (60, 150) ms | Hold duration on buttons |
| click_input_x_range | (0.05, 0.30) | Click position within input (left side) |
Idle Behavior
| Parameter | Default | Effect |
|---|---|---|
| idle_between_actions | False | Enable micro-movements between actions |
| idle_between_duration | (0.3, 0.8) sec | Idle duration between actions |
| idle_drift_px | 3 px | Random cursor drift during idle |
| idle_pause_range | (300, 1000) ms | Idle pause duration |
| initial_cursor_x | (400, 700) px | Starting cursor X (address bar area) |
| initial_cursor_y | (45, 60) px | Starting cursor Y (address bar area) |
Presets
| Preset | Typing | Aim Delay | Idle Between | Use Case |
|---|---|---|---|---|
| default | 70±40ms | 60-140ms | No | Normal human speed |
| careful | 100±50ms | 80-180ms | Yes (0.4-1.0s) | Slower, more deliberate |
Pitfalls
- Humanize requires the wrapper. Connecting via CDP without the wrapper loses humanization — only fingerprint patches work over raw CDP.
- ElementHandle objects bypass humanization in Playwright. Use
page.click(selector)orpage.locator(selector).*— avoidquery_selector()handles. page.fill()with humanize clears existing content and types character by character. This is intentional but slower than rawfill().- Typing speed profiling exists. Some sites measure ms-between-keystrokes. Use
carefulpreset or increasetyping_delayfor suspicious targets. - Scroll-to-element is automatic. On click/hover, the element is scrolled into view with human scroll before interaction.
- CDP Isolated Worlds are used for stealth DOM queries. This prevents monkey-patch detection in the main JavaScript context.
Verification
- Test against
https://deviceandbrowserinfo.com— should show 24/24 behavioral signals passed. - Compare reCAPTCHA v3 scores with and without
humanize=True— humanize should increase score by 0.2-0.4. - Use browser DevTools Performance tab to record a session — mouse events should show realistic timing curves, not instant jumps.
- Verify
Input.dispatchKeyEventevents haveisTrusted: truein the event listener.
Related Skills
stealth-browser-launch— Launch the patched Chromium binary with C++ fingerprint modifications.tls-fingerprint-impersonation— TLS/JA4 fingerprint spoofing for HTTP-level evasion.
Related Skills
Agent-Reach
112.3kGive your AI agent eyes to see the entire internet. Read & search Twitter, Reddit, YouTube, GitHub, Bilibili, XiaoHongShu — one CLI, zero API fees.
ruflo
81.9k🌊 The original agent harness. Deploy intelligent multi-player swarms, coordinate autonomous workflows, and build conversational AI systems. Features adaptive memory, self-learning intelligence, federation, vector RAG integration, and native Claude Code / Codex / Hermes and many more Integrated
headroom
79.8kCompress tool outputs, logs, files, and RAG chunks before they reach the LLM. 20% fewer tokens for coding agents, 60-95% fewer tokens for JSON, same answers. Library, proxy, MCP server.
Claude-Code-Game-Studios
39.7kTurn Claude Code into a full game dev studio — 49 AI agents, 72 workflow skills, and a complete coordination system mirroring real studio hierarchy.
Languages
Trust signals
From repository metadata: license, adoption, age and documentation. Not a code audit — see the Safety scan above for what the skill file itself contains.
