android-deeplink-applink-abuse
Exploit Android deep links and App Links — unvalidated params in BROWSABLE activities, custom-scheme OAuth-callback hijacking, open redirect into WebViews, and App Links verification gaps (missing/invalid assetlinks.json). Tier 1. Use when the manifest declares deep-link intent-filters. Aligns to OWASP MASTG MASVS-PLATFORM (MASTG-TEST-0028).
AnandJogawade · AI review on file
Installation
Read the source and dependencies before running a command. Copying does not install or execute anything.
npx skills add AnandJogawade/Zazy-Pentest-Plugin-for-Claude --skill android-deeplink-applink-abuseAbout this skill
name: android-deeplink-applink-abuse description: Exploit Android deep links and App Links — unvalidated params in BROWSABLE activities, custom-scheme OAuth-callback hijacking, open redirect into WebViews, and App Links verification gaps (missing/invalid assetlinks.json). Tier 1. Use when the manifest declares deep-link intent-filters. Aligns to OWASP MASTG MASVS-PLATFORM (MASTG-TEST-0028).
Android Deep Link & App Link Abuse
Deep links are a remote, one-click entry point. Fuzz their params, hijack custom-scheme callbacks, and check whether App Links are actually verified.
Trigger phrases
deep link, applink, app link, browsable, custom scheme, assetlinks.json, autoverify, oauth callback hijack, deeplink open redirect
Method
- Enumerate schemes/hosts from the manifest (
<data scheme/host>+BROWSABLE):drozer> run app.package.manifest <package>
- Fire deep links & fuzz params:
adb shell am start -a android.intent.action.VIEW -d "myapp://host/path?param=value"drozer> run app.activity.start --action android.intent.action.VIEW --data-uri "myapp://host/auth=sum"- Reach authenticated/internal screens unauthenticated; inject into params that feed a WebView or backend call.
- Path traversal in params:
adb shell am start -a android.intent.action.VIEW -d "example://app/users?u=../../unwanted%3fp=v" - App Links verification:
curl -s https://<domain>/.well-known/assetlinks.json— missing/invalid SHA-256 fingerprint →autoVerifysilently falls back to the browser / chooser (not enforced), enabling scheme hijack.
Modern / rare techniques
- [NEW/RARE] Custom-scheme OAuth-callback hijacking — any app can register
myapp://; a malicious app intercepting the OAuth redirect steals the auth code/token → one-click account takeover. PKCE mitigates — note whether it's used. - [NEW/RARE] Open redirect via weak
startsWith/containshost checks feeding a WebView → load attacker page (chainandroid-webview-attacks). - [NEW/RARE] intent:// redirection from a BROWSABLE activity to reach non-exported components (chain
android-intent-injection-redirection). - [NEW/RARE] Expired allowlisted domain re-registered by an attacker receives auth headers/tokens.
Bypass / evasion
Defeat startsWith/contains/suffix host allowlists with attacker.com#victim.com, victim.com.attacker.com, //attacker.com, and userinfo (victim.com@attacker.com) tricks before the WebView load.
Escalation / chaining
One-click ATO (token theft), LFI/XSS-to-native via WebView, open redirect. Chains to android-webview-attacks, android-intent-injection-redirection, oauth, and android-backend-api-testing (confirm server accepts the stolen token).
Backend & tooling
Preferred Kali path (native Kali / WSL+Kali / Kali-SSH / Kali-MCP / HexStrike). Needs the user's device/emulator (adb/drozer; a PoC app for scheme-hijack) + curl for assetlinks. Route token tests through Burp if a Burp MCP is present. Windows-local last resort. See zazy-orchestrator.
Verification (background, before reporting)
Confirm the token/session actually leaked to an attacker-controlled URI or the non-exported component truly launched — use OAST/a second app to receive the callback — before marking CONFIRMED. For "unverified" App Links, show the chooser/browser fallback actually occurs.
Evidence
The deep-link URI, the manifest intent-filter, the leaked token/reached screen, and (for App Links) the missing/invalid assetlinks.json. Remediation: validate all params; verified App Links + PKCE for OAuth; canonical host allowlists; reject nested/implicit intents.