← 返回 Skills 市场
sarperarikan

AccessMind

作者 sarperarikan · GitHub ↗ · v6.0.0 · MIT-0
cross-platform ⚠ suspicious
152
总下载
0
收藏
1
当前安装
14
版本数
在 OpenClaw 中安装
/install accessmind
功能描述
Enterprise WCAG 2.2/2.1/EN 301 549 accessibility audit platform. OpenClaw Browser Tool ile entegre çalışan, güvenli ve profesyonel erişilebilirlik denetim si...
使用说明 (SKILL.md)

AccessMind Enterprise v6.0

🆕 v6.0: Güvenli ve Profesyonel Erişilebilirlik Denetimi

AccessMind, OpenClaw Browser Tool ile tam entegre çalışan, güvenli ve profesyonel erişilebilirlik denetim platformudur.

┌─────────────────────────────────────────────────────────────┐
│  AccessMind Enterprise v6.0                                  │
├─────────────────────────────────────────────────────────────┤
│  1. OpenClaw Browser → Güvenli site açma                    │
│  2. ARIA Snapshot → DOM analizi                             │
│  3. Keyboard Navigation → Tab simülasyonu                   │
│  4. Focus Testing → Focus trap tespiti                      │
│  5. Visual Analysis → Screenshot + LLM analiz                │
│  6. ACT Rules → 50+ WCAG kuralı                             │
│  7. Professional Report → HTML/JSON çıktı                   │
└─────────────────────────────────────────────────────────────┘

🛡️ Güvenlik

AccessMind v6.0, tüm tarayıcı işlemlerini OpenClaw Browser Tool üzerinden gerçekleştirir. Bu sayede:

  • Güvenli tarama - OpenClaw yönetilen tarayıcı kullanır
  • Cloudflare uyumlu - Browser tool, korumalı sitelerde çalışır
  • İzolü ortam - Sandbox içinde çalışır
  • Şeffaf - Tüm işlemler loglanır
  • Hafif - Ekstra browser kurulumu gerektirmez

🚀 Hızlı Başlangıç

Örnek Kullanım

Kullanıcı: https://arcelik.com.tr için erişilebilirlik denetimi yap

AccessMind:
1. Browser tool ile siteyi açar
2. ARIA snapshot alır
3. Klavye navigasyon testi yapar
4. Focus trap kontrolü yapar
5. Ekran görüntüsü alır
6. ACT Rules çalıştırır
7. Profesyonel rapor oluşturur

OpenClaw Browser Tool ile Entegrasyon

# Siteyi aç
browser action=open url=https://example.com

# ARIA snapshot al (erişilebilirlik için optimize)
browser action=snapshot refs=aria

# Klavye navigasyonu test et
browser action=act kind=press key=Tab
browser action=snapshot refs=aria

# Focus kontrolü
browser action=act kind=press key=Tab
browser action=snapshot refs=aria

# Ekran görüntüsü al
browser action=screenshot

⌨️ Klavye Navigasyon Testi

AccessMind, gerçek kullanıcı davranışını simüle eder:

  1. Tab Simülasyonu - Tüm focusable elementleri gez
  2. Focus Tracking - Focus değişimlerini takip et
  3. Focus Indicator - Focus göstergesi kontrolü
  4. Focus Trap Detection - Klavye tuzaklarını tespit et
  5. Escape Test - Modal ve dialog kapatma testi

Sonuç Formatı

{
  "keyboard_navigation": {
    "total_steps": 50,
    "focus_changes": 31,
    "visible_outlines": 28,
    "focus_visible_support": 25,
    "issues": 12
  },
  "focus_traps": {
    "total": 0,
    "details": []
  },
  "metrics": {
    "focus_efficiency": 85,
    "keyboard_accessibility": 76,
    "sr_friendliness": 92
  }
}

📊 ACT Rules (50+ Kural)

AccessMind, W3C ACT Rules Format'a uygun 50+ atomik kural içerir:

Perceivable (Algılanabilir)

Kural ID Kriter Açıklama
ACT-1.1 1.1.1 Image has accessible name
ACT-1.2 1.1.1 SVG has accessible name
ACT-1.3 1.1.1 Area has accessible name
ACT-2.1 1.3.1 Heading has content
ACT-2.2 1.3.1 List has proper structure
ACT-2.3 1.3.1 Table has headers
ACT-3.1 1.4.3 Text has sufficient contrast
ACT-3.2 1.4.3 UI components have sufficient contrast
ACT-3.3 1.4.4 Text resizes properly
ACT-3.4 1.4.10 Content reflows horizontally

Operable (İşletilebilir)

Kural ID Kriter Açıklama
ACT-4.1 2.1.1 Interactive element is keyboard accessible
ACT-4.2 2.1.1 Focusable element has keyboard event
ACT-4.3 2.1.2 No keyboard trap
ACT-4.4 2.4.1 Page has bypass blocks
ACT-4.5 2.4.2 Page has title
ACT-4.6 2.4.3 Focus order is logical
ACT-4.7 2.4.4 Link has accessible name
ACT-4.8 2.4.4 Link purpose is clear
ACT-4.9 2.4.6 Heading describes topic
ACT-4.10 2.4.7 Focus is visible
ACT-4.11 2.5.1 Clickable has accessible name
ACT-4.12 2.5.5 Target size is sufficient

Understandable (Anlaşılabilir)

Kural ID Kriter Açıklama
ACT-5.1 3.1.1 Page language is specified
ACT-5.2 3.1.2 Part language is specified
ACT-5.3 3.2.1 Focus doesn't change context
ACT-5.4 3.2.2 Input doesn't change context unexpectedly
ACT-5.5 3.2.4 Consistent navigation
ACT-5.6 3.3.1 Error identification
ACT-5.7 3.3.2 Form has labels
ACT-5.8 3.3.3 Error suggestion
ACT-5.9 3.3.4 Error prevention

Robust (Sağlam)

Kural ID Kriter Açıklama
ACT-6.1 4.1.1 Parsing is valid
ACT-6.2 4.1.2 Custom element has accessible name
ACT-6.3 4.1.2 Custom element has role
ACT-6.4 4.1.2 ARIA state is valid
ACT-6.5 4.1.2 Form has accessible name
ACT-6.6 4.1.3 Status message has role

📈 Innovative Metrics

Focus Efficiency Score (0-100)

Focus'un sayfada ne kadar verimli hareket ettiğini ölçer:

  • Focus değişim tutarlılığı (40%)
  • Visible focus outlines (30%)
  • :focus-visible desteği (30%)

Keyboard Accessibility Score (0-100)

Klavye erişilebilirlik sorunlarını ölçer:

  • Critical sorun: -10 puan
  • Serious sorun: -5 puan
  • Moderate sorun: -2 puan

Screen Reader Friendliness Score (0-100)

ARIA live region kalitesini ölçer:

  • Doğru aria-live attribute'ları (50%)
  • Dinamik içerik duyuruları (50%)

Focus Trap Risk Score (0-100)

Klavye tuzağı riskini ölçer:

  • 0 tuzak: 100 puan
  • 1 tuzak: 70 puan
  • 2 tuzak: 40 puan
  • 3+ tuzak: 20 puan

📝 Profesyonel Raporlama

HTML Rapor

\x3C!DOCTYPE html>
\x3Chtml lang="tr">
\x3Chead>
    \x3Ctitle>WCAG Erişilebilirlik Raporu - example.com\x3C/title>
    \x3Cmeta charset="UTF-8">
    \x3Cmeta name="viewport" content="width=device-width, initial-scale=1.0">
\x3C/head>
\x3Cbody>
    \x3Cheader>
        \x3Ch1>Erişilebilirlik Denetim Raporu\x3C/h1>
        \x3Cp>Site: example.com\x3C/p>
        \x3Cp>Tarih: 2026-03-26\x3C/p>
    \x3C/header>
    
    \x3Csection id="summary">
        \x3Ch2>Özet\x3C/h2>
        \x3Cdiv class="score">Overall Score: 85/100\x3C/div>
        \x3Cdiv class="issues">Toplam İhlal: 12\x3C/div>
    \x3C/section>
    
    \x3Csection id="wcag-criteria">
        \x3Ch2>WCAG Kriterleri\x3C/h2>
        \x3C!-- Her kriter için detaylı sonuç -->
    \x3C/section>
    
    \x3Csection id="recommendations">
        \x3Ch2>Öneriler\x3C/h2>
        \x3C!-- Düzeltme önerileri -->
    \x3C/section>
\x3C/body>
\x3C/html>

JSON Rapor

{
  "audit_info": {
    "url": "https://example.com",
    "date": "2026-03-26T08:00:00Z",
    "auditor": "AccessMind Enterprise v6.0",
    "standard": "WCAG 2.2 AA"
  },
  "summary": {
    "overall_score": 85,
    "total_issues": 12,
    "critical": 2,
    "serious": 5,
    "moderate": 5
  },
  "criteria_results": [
    {
      "criterion": "1.1.1",
      "name": "Non-text Content",
      "result": "FAIL",
      "issues": [
        {
          "element": "\x3Cimg src='logo.png'>",
          "issue": "Image missing alt attribute",
          "severity": "critical",
          "wcag": "1.1.1"
        }
      ]
    }
  ],
  "metrics": {
    "focus_efficiency": 85,
    "keyboard_accessibility": 76,
    "sr_friendliness": 92,
    "focus_trap_risk": 100
  }
}

🔧 Kullanım

CLI (Eski Script'ler)

# Klavye navigasyon testi
python3 /Users/sarper/.openclaw/workspace/skills/accessmind/scripts/accessmind-behavioral-navigator.py \
  --url https://example.com \
  --steps 50 \
  --output /Users/sarper/.openclaw/workspace/audits

# Derinlemesine tarama
python3 /Users/sarper/.openclaw/workspace/skills/accessmind/scripts/accessmind-deep-crawler.py \
  --url https://example.com \
  --depth 5 \
  --pages 12 \
  --output /Users/sarper/.openclaw/workspace/audits

OpenClaw Browser Tool (Önerilen)

AccessMind artık OpenClaw Browser Tool ile tam entegre çalışıyor. Kullanıcıdan gelen "siteyi denetle" talebi otomatik olarak browser tool ile gerçekleştirilir.


📁 Dosya Yapısı

/Users/sarper/.openclaw/workspace/skills/accessmind/
├── SKILL.md                              # Ana dokümantasyon
├── references/
│   ├── wcag-2.2-criteria.md              # WCAG 2.2 kriterleri
│   ├── aria-guide.md                     # ARIA kılavuzu
│   ├── screen-reader-patterns.md         # Screen reader kalıpları
│   ├── voiceover-patterns.md             # VoiceOver kalıpları
│   ├── common-issues.md                  # Yaygın sorunlar
│   ├── browser-tool-workflow.md          # Browser tool workflow
│   └── mobile-testing.md                 # Mobil test
├── scripts/
│   ├── accessmind-behavioral-navigator.py # Klavye navigasyon testi
│   ├── accessmind-deep-crawler.py         # Derinlemesine tarama
│   ├── act-rules-engine.py                # ACT Rules motoru
│   ├── report-generator.py                # Rapor oluşturucu
│   └── wcag-em-evaluator.py               # WCAG-EM değerlendirici
└── assets/
    └── report-template.md                 # Rapor şablonu

🎯 WCAG Kriterleri

Kriter Kategori Test Edildi
1.1.1 Non-text Content
1.2.1 Audio-only and Video-only
1.3.1 Info and Relationships
1.3.2 Meaningful Sequence
1.4.3 Contrast Minimum
1.4.4 Resize Text
1.4.10 Reflow
1.4.11 Non-text Contrast
2.1.1 Keyboard
2.1.2 No Keyboard Trap
2.1.4 Character Key Shortcuts
2.4.1 Bypass Blocks
2.4.2 Page Titled
2.4.3 Focus Order
2.4.4 Link Purpose
2.4.5 Multiple Ways
2.4.6 Headings and Labels
2.4.7 Focus Visible
2.5.1 Pointer Gestures
2.5.2 Pointer Cancellation
2.5.3 Label in Name
2.5.4 Motion Actuation
2.5.5 Target Size
2.5.8 Target Size (Minimum)
3.1.1 Language of Page
3.1.2 Language of Parts
3.2.1 On Focus
3.2.2 On Input
3.2.3 Consistent Navigation
3.2.4 Consistent Identification
3.3.1 Error Identification
3.3.2 Labels or Instructions
3.3.3 Error Suggestion
3.3.4 Error Prevention
3.3.7 Redundant Entry
4.1.2 Name, Role, Value
4.1.3 Status Messages

📚 Referanslar


AccessMind Enterprise v6.0
Güvenli ve Profesyonel Erişilebilirlik Denetimi
OpenClaw Browser Tool Entegrasyonu
WCAG 2.2 AA Uyumluluk

安全使用建议
What to check before installing: - Inspect the hard-coded path (/Users/sarper/.openclaw/...) in extension/background.js; it should be replaced with a configurable path or discovered dynamically. Hard-coded user paths can cause failures or accidentally point to sensitive files. - Expect the extension to collect full page HTML, screenshots, images, and form element metadata (names/labels). If you will run audits on pages that contain sensitive data, be cautious — exported audit artifacts could include sensitive content. - The skill talks to local services (OpenClaw Gateway at 127.0.0.1:8765 and an Ollama-like API at 127.0.0.1:11434). Verify you are comfortable running those local services and that they are configured to accept only local connections. - Review the included shell scripts (.sh) and Python scripts before running them. They perform crawling and reporting and may invoke system commands or write files. Prefer to run them in a controlled environment (isolated VM or container) first. - Confirm whether the extension is the only component that will execute automatically; the repository contains many scripts that likely require manual execution. If you do not want background local servers or files written to disk, do not run those scripts. - If you intend to use this on enterprise or sensitive sites, get written assurances (or source review) that exported reports and stored event logs will not be transmitted to remote servers. The code appears to use only localhost endpoints, which is better than unknown remote endpoints, but verify at runtime. If you want, I can: (1) point to the exact lines that reference the hard-coded path and localhost endpoints; (2) summarize what each .sh and top Python script does so you can decide which to allow; or (3) suggest a minimal, safer configuration for running audits (e.g., run in a disposable VM and disable export/auto-tracking).
能力评估
Purpose & Capability
Name, description, SKILL.md and the included files (content/background/devtools scripts, many Python audit scripts) are consistent with an accessibility auditing tool. The declared dependency on python3 and Python parsing packages aligns with the presence of many Python scripts. However, the code contains a hard-coded absolute path to a user's workspace (/Users/sarper/.openclaw/...), which is unusual for a general-purpose skill and suggests the package was packaged from a specific developer environment without path generalization.
Instruction Scope
Content/background scripts gather full page HTML, images, element attributes, form inputs and take screenshots; they also store behavioral event logs in chrome.storage.local and can export them. While this data collection is plausible for an auditor, it can capture sensitive information (page HTML, input names/values, screenshots). SKILL.md does not explicitly warn about collecting page HTML/screenshots or export behavior; background.js also attempts to send images and URLs to local endpoints (OpenClaw Gateway / Ollama). The presence of many shell scripts (.sh) and Python scripts that appear to implement crawling and auditing means the skill can perform network requests, write output files, and run system commands if executed—this expands the runtime surface beyond the simple browser actions shown in SKILL.md.
Install Mechanism
There is no install spec (instruction-only), which reduces automated installation risk. However, the repo includes many executable artifacts (multiple .py and .sh scripts) and a browser extension. Those files will be written to disk when the skill is installed by the platform and could be executed later. No external download URLs are used (good), but the presence of shell scripts (audit-page.sh, full-audit.sh, stealth-audit.sh) and scripts that may spawn processes means users should review them before running. The background script also expects local services (OpenClaw Gateway, Ollama) to be present.
Credentials
The skill declares python3 and Python parsing libraries (beautifulsoup4, lxml, html5lib) — reasonable for web crawling and HTML analysis. It does not request environment variables, cloud credentials, or unrelated secrets. The only surprising element is reliance on local services (localhost:8765, localhost:11434) and a hard-coded developer workspace path; these are not environment variables but are environment-dependent and may fail or leak data if pointed incorrectly.
Persistence & Privilege
always is false (normal). The Chrome extension code stores behavioralEvents in chrome.storage.local (up to 1000 events) and exposes an export function; it also logs operations and may create reports on disk via Python scripts. That is expected for an audit tool, but it means persistent storage of potentially sensitive page data. The skill does not declare modifying other skills or system-wide configurations, but the hard-coded absolute paths and included shell scripts mean a careful review of what those scripts write or execute is recommended.
如何使用
  1. 确保已安装 OpenClaw(本地或 Docker 部署)
  2. 在对话框中输入安装命令:/install accessmind
  3. 安装完成后,直接呼叫该 Skill 的名称或使用 /accessmind 触发
  4. 根据 Skill 的参数说明提供必要输入,即可获得结构化输出
版本历史
v6.0.0
v6.0: Security improvements, 56 ACT Rules, OpenClaw Browser Tool integration, Professional HTML/JSON reports
v5.1.0
v5.1: OpenClaw browser tool integration, Cloudflare bypass, real-time accessibility testing, improved workflow documentation
v5.0.0
v5.0: Agentic keyboard navigation, focus trap detection, ARIA live region testing, dynamic content analysis, innovative metrics
v3.6.0
v3.6.0: Interactive Element Testing & Keyboard Simulation added. Element-by-element traversal from top to bottom, ARIA attribute analysis (missing/invalid ARIA), JavaScript event detection (click without keyboard), CSS accessibility checks (outline removal, fixed font sizes), HTML semantic validation (labels, alt text, heading hierarchy), Keyboard simulation with Python keyboard library (Tab, Enter, Arrow, Escape), pytest integration for CI/CD. Tested on Arcelik.com.tr - 3,543 elements analyzed, 141 issues found.
v3.5.1
v3.5.1: Default LLM changed to GLM 5 Cloud (glm-5:cloud) for better reasoning. Cloudflare Bypass Browser Automation - Real Chrome non-headless bypasses bot detection, Human behavior simulation, Stealth mode, DevTools inspection. Tested on Beko.com.tr - 11 violations detected.
v3.5.0
v3.5: Cloudflare Bypass Browser Automation added - Real Chrome (non-headless) bypasses bot detection, Human behavior simulation (mouse/scroll/random delays), Stealth mode removes automation flags, DevTools element inspection + screenshot capture, Tested on Beko.com.tr - 11 violations detected. Also includes LangGraph multi-agent workflow, Ollama integration (qwen3.5:cloud), Dual-engine audit (Axe-core + Pa11y), Auto-trigger capability.
v1.0.7
### Version 1.0.7 Changelog - Bumped internal metadata version from 1.0.6 to 1.1.0 in SKILL.md. - No changes to the functionality, capabilities, or usage instructions. - Documentation update only; core features and standards support remain unchanged.
v1.0.6
- Updated metadata version to 1.0.6. - No feature changes; documentation and functionality remain unchanged.
v1.0.5
- Updated to version 1.0.5. - Internal metadata version bumped from 3.3.0 to 1.0.5. - Removed redundant/replaced file: _meta.json. - No changes to feature set or user-facing instructions.
v1.0.4
- Updated engine to AccessMind Enterprise v3.3.0 (from 3.2.0). - No changes to features, API, or usage. - Updated internal version in SKILL.md for platform metadata consistency.
v1.0.3
- Updated Openclaw metadata version from 3.1.0 to 3.2.0. - No functional or documentation changes to features, usage, or descriptions. - Ensures compatibility and clarity for dependency management.
v1.0.2
- English translation and terminology improvements in all descriptions and usage instructions. - Metadata updated to include dual testing engine support (axe-core and pa11y). - Capability and usage sections revised for clarity and English language consistency. - Version and platform references updated to v3.1 and to reflect W3C standards terminology. - Removal of Turkish content; all documentation now in English.
v1.0.1
**AccessMind 1.0.1 — Adds dual-engine automated accessibility testing.** - Added Pa11y testing engine alongside Axe-core for enhanced audit reliability. - Updated requirements and install steps to include the pa11y npm package. - Introduced dual-engine and violation deduplication support in metadata. - Documentation now reflects new engine setup, versions, and feature capabilities.
v1.0.0
Initial release of AccessMind: Enterprise-grade accessibility audit platform, fully aligned with W3C standards. - Implements full WCAG-EM 5-step methodology for website accessibility evaluation. - Includes ACT Rules (22 atomic W3C-format rules) for automated assessment. - Simulates Combined Expertise evaluation with 5 distinct roles. - Tests accessibility via 5 disability profiles/user journeys for real-world validation. - Supports LLM-powered human-centric reporting and narrative summaries. - Exports results in EARL (W3C RDF) format for interoperability. - Provides both automated and manual testing workflows for comprehensive audits.
元数据
Slug accessmind
版本 6.0.0
许可证 MIT-0
累计安装 1
当前安装数 1
历史版本数 14
常见问题

AccessMind 是什么?

Enterprise WCAG 2.2/2.1/EN 301 549 accessibility audit platform. OpenClaw Browser Tool ile entegre çalışan, güvenli ve profesyonel erişilebilirlik denetim si... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 152 次。

如何安装 AccessMind?

在 OpenClaw 或 Claude Code 对话框中运行命令「/install accessmind」即可一键安装,无需额外配置。

AccessMind 是免费的吗?

是的,AccessMind 完全免费,采用 MIT-0 许可证,可自由下载、安装和使用。

AccessMind 支持哪些平台?

AccessMind 跨平台运行,可在任意部署了 OpenClaw / Claude Code 的环境中使用(cross-platform)。

谁开发了 AccessMind?

由 sarperarikan(@sarperarikan)开发并维护,当前版本 v6.0.0。

💬 留言讨论