CommunityResearch & Data Analysisgithub.com

Cyarun/website-dpia-audit

Use when conducting a read-only, reputation-safe Website Data Protection Impact Assessment (DPIA) & Client-Side Automation Readiness Audit under India's Digital Personal Data Protection Act, 2023 (DPDP Act) and DPDP Rules, 2025 (G.S.R. 846(E)). Detects pre-consent cookies, storage, marketing/CRM automations, CDPs, session replay, AI chatbots, ad pixels, dark patterns, security headers, 90-day grievance SLA signals, and cross-border sub-processors across Claude, Gemini, DeepSeek, and Qwen.

What is website-dpia-audit?

website-dpia-audit is a Claude Code agent skill that use when conducting a read-only, reputation-safe Website Data Protection Impact Assessment (DPIA) & Client-Side Automation Readiness Audit under India's Digital Personal Data Protection Act, 2023 (DPDP Act) and DPDP Rules, 2025 (G.S.R. 846(E)). Detects pre-consent cookies, storage, marketing/CRM automations, CDPs, session replay, AI chatbots, ad pixels, dark patterns, security headers, 90-day grievance SLA signals, and cross-border sub-processors across Claude, Gemini, DeepSeek, and Qwen.

Works with✓Claude Code~Codex CLI~Cursor✓Gemini CLI
npx skills add https://github.com/Cyarun/dpdpa-website-dpia-skill/tree/HEAD/.agents/skills/website-dpia-audit

Installed? Explore more Research & Data Analysis skills: affaan-m/uspto-database, affaan-m/scholar-evaluation, affaan-m/literature-review · View all 6 →

Ask in your favorite AI

Open a new chat with this agent skill pre-loaded.

Documentation

Website DPIA & Automation Readiness Auditor (website-dpia-audit)

Purpose & Scope (Websites Only — Read-Only & Reputation-Safe)

This skill conducts an empirical, read-only, non-intrusive Website DPIA (Data Protection Impact Assessment) & Automation Readiness Assessment for any target website URL or local HTML build. It is engineered to work deterministically across Claude (3.5/3.7/4 Sonnet & Opus), Google Gemini (2.5 Pro/Flash), DeepSeek (R1 & V3), and Alibaba Qwen (Qwen-2.5/Qwen3/QwQ).


Hard Invariants (Never Violate)

  1. Statutory Precision (Tier-1 Canonical Sources Only):

    • DPDP Act, 2023 (Act No. 22 of 2023): 9 Chapters, 44 Sections, 1 Schedule.
    • DPDP Rules, 2025 (G.S.R. 846(E), notified 13 November 2025): 23 Rules, 7 Schedules.
      • Phased Commencement: 13 Nov 2025 (Rules 1–2, 17–21); 13 Nov 2026 (+1 yr: Rule 4 Consent Manager registration); 13 May 2027 (+18 mo: Rules 3, 5–16, 22–23 operational compliance).
    • Grievance Redressal SLA: Maximum 90 days under Rule 14(3) read with Section 13 (never cite 30 or 7 days as the DPDP statutory cap).
    • Security & Access Log Retention: Minimum 1 year under Rule 8(3) read with the Seventh Schedule.
    • DPIA & Algorithmic Risk Assessment Mandate: Section 10(2)(c)(ii) read with Rule 13(1)–(3) (annual DPIA for Significant Data Fiduciaries + duty to verify that algorithmic software, AI chatbots, session replay, and automated profiling do not pose a risk to the rights of Data Principals).
    • Children's Data: Section 9(1)–(3) read with Rules 10–11 and the Fourth Schedule (verifiable parental consent for under-18 users; strict prohibition on tracking, behavioural monitoring, and targeted advertising directed at children).
  2. Reputation-Safe, Constructive Advisory Framing (Zero Defamation):

    • Every report MUST be titled and framed as a "Confidential & Constructive Advisory DPIA Readiness Assessment" based solely on publicly observable client-side signals at the timestamp of inspection.
    • Prohibited Words: Never use illegal, unlawful, lawbreaker, violator, criminal, guilty, penalized, fineable, sued, offender, scam, or fraud.
    • Required 5-Part Finding Structure:
      1. observed_technical_signal (exact DOM/header/cookie/network evidence; zero hallucination)
      2. statutory_mapping (exact DPDP Act 2023 Section + DPDP Rules 2025 Rule)
      3. readiness_tier (Aligned | Advisory Enhancement | Priority Readiness Action)
      4. reputation_safe_advisory_note (neutral, board-ready explanation)
      5. recommended_engineering_fix (concrete code/header/CMP configuration step)

Step-by-Step Audit Workflow

Step 1: Collect Empirical Website Telemetry & Baseline DPIA Report

Run the read-only telemetry collector against the target website URL (or local HTML file):

node products/dpia-audit-skill/scripts/collect_website_dpia_telemetry.mjs "<https://target-website.example>" \
  --out "/tmp/dpia-telemetry-report.json" \
  --model "<claude|gemini|deepseek|qwen|deterministic-telemetry-only>"

What collect_website_dpia_telemetry.mjs automatically inspects in read-only mode:

  • A. Pre-Consent Storage & Cookies: First-party and third-party cookies set prior to any user interaction (httpOnly, secure, sameSite, expiry), localStorage, sessionStorage, IndexedDB, and ServiceWorker registrations.
  • B. Website Automations, AI Agents, Session Replay & Behavioral Profiling (Rule 13(2) & Section 6):
    • Marketing & CRM Automations: HubSpot, Marketo, Pardot, ActiveCampaign, Klaviyo, Braze, MoEngage, WebEngage, CleverTap, Zoho PageSense/SalesIQ, Mailchimp, Customer.io, Drip, OmniSend.
    • CDP & Event Routers: Segment (analytics.js), Google Tag Manager (gtm.js), RudderStack, mParticle, Snowplow, Tealium, Adobe Launch.
    • Session Replay & Behavioral Heatmaps: Microsoft Clarity, Hotjar, FullStory, LogRocket, Smartlook, PostHog, Mouseflow, CrazyEgg, Contentsquare, LuckyOrange.
    • AI Chatbots & Support Widgets: Intercom, Drift, Crisp, Freshchat, Zendesk, Tidio, LiveChat, Botpress, Voiceflow, Ada, Yellow.ai, Haptik, Verloop, Gupshup, Wati / WhatsApp Business floating widgets.
    • Ad Pixels & Auto-Advanced Matching: Meta Pixel (fbevents.js, fbq), Google Ads/Analytics (gtag.js), LinkedIn Insight (lintrk), X/Twitter, Snap, TikTok, Criteo, Taboola, Outbrain.
    • Form Listeners & Dark Patterns: Pre-ticked checkboxes (input[type="checkbox"][checked]), bundled consent, insecure http:// form actions, pre-submit field capture listeners, on-load Web Push prompts (Notification.requestPermission), and Canvas/Audio fingerprinting signals.
  • C. Statutory Control Pillars:
    • Sec 5 / Rule 3: Privacy Notice link presence & Eighth Schedule Indian language availability signals.
    • Sec 6 / Rules 3–4: CMP presence (OneTrust, Cookiebot, CookieYes, Osano, TrustArc, Securiti, Didomi, Klaro/custom) & visual symmetry between Accept and Reject / Essential Only controls + consent withdrawal link.
    • Sec 8(5) / Rule 6 / Rule 8(3): HTTPS, HSTS, CSP, X-Content-Type-Options, Referrer-Policy, Permissions-Policy, Subresource Integrity (integrity=) on external scripts, and 1-year log retention advisory.
    • Sec 9 / Rules 10–11: Children's audience signals vs. active behavioural/ad trackers.
    • Sec 13 / Rule 14(3): DPO / Grievance Officer contact details and 90-day statutory SLA verification.
    • Sec 16 / Rule 15: Cross-border sub-processor domain inventory.

Step 2: Synthesize & Validate Against the Multi-Model Schema

  1. Load the universal prompt pack from products/dpia-audit-skill/prompts/universal-dpia-system-prompt.md.
  2. Validate the output JSON against products/dpia-audit-skill/schema/dpia-website-report.schema.json.
  3. Present both the structured JSON report and an executive Markdown summary table for the user, preserving the Confidential & Constructive Advisory DPIA Readiness Assessment disclaimer.

Reference Artifacts

  • Telemetry Collector & Report Builder: products/dpia-audit-skill/scripts/collect_website_dpia_telemetry.mjs
  • Multi-Model Prompt Pack (Claude / Gemini / DeepSeek / Qwen): products/dpia-audit-skill/prompts/universal-dpia-system-prompt.md
  • JSON Schema Contract: products/dpia-audit-skill/schema/dpia-website-report.schema.json

Related Skills