Android Reverse Engineering

Android Reverse Engineering

Decompile Android APK, XAPK, AAB, DEX, JAR, and AAR files using jadx and Fernflower/Vineflower, trace call flows through application code and libraries, analyze security patterns, produce structured documentation of extracted APIs, and perform adaptive dynamic analysis with Frida — generating custom bypass scripts based on what the static analysis finds, iterating through crash logs to refine hooks until protections are bypassed. Two decompiler engines are supported — jadx for broad Android coverage and Fernflower for higher-quality output on complex Java code — and can be used together for comparison.

Prerequisites

This skill requires Java JDK 17+ and jadx to be installed. Fernflower/Vineflower and dex2jar are optional but recommended for better decompilation quality. bundletool is required for AAB (App Bundle) files. For dynamic analysis (Phase 7), Python 3.8+, adb, and a device/emulator with frida-server are needed — the setup-frida.sh script handles the full setup. Run the dependency checker to verify:

bash ${CLAUDE_PLUGIN_ROOT}/skills/android-reverse-engineering/scripts/check-deps.sh

If anything is missing, follow the installation instructions in ${CLAUDE_PLUGIN_ROOT}/skills/android-reverse-engineering/references/setup-guide.md.

Workflow

Phase 1: Verify and Install Dependencies

Before decompiling, confirm that the required tools are available — and install any that are missing.

Action: Run the dependency check script.

bash ${CLAUDE_PLUGIN_ROOT}/skills/android-reverse-engineering/scripts/check-deps.sh

The output contains machine-readable lines:

• INSTALL_REQUIRED:<dep> — must be installed before proceeding
• INSTALL_OPTIONAL:<dep> — recommended but not blocking

If required dependencies are missing (exit code 1), install them automatically:

bash ${CLAUDE_PLUGIN_ROOT}/skills/android-reverse-engineering/scripts/install-dep.sh <dep>

The install script detects the OS and package manager, then:

• Installs without sudo when possible (downloads to ~/.local/share/, symlinks in ~/.local/bin/)
• Uses sudo and the system package manager when necessary (apt, dnf, pacman)
• If sudo is needed but unavailable or the user declines, it prints the exact manual command and exits with code 2 — show these instructions to the user

For optional dependencies, ask the user if they want to install them. Vineflower and dex2jar are recommended for best results.

After installation, re-run check-deps.sh to confirm everything is in place. Do not proceed to Phase 2 until all required dependencies are OK.

Phase 2: Decompile

Use the decompile wrapper script to process the target file. The script supports three engines: jadx, fernflower, and both.

Action: Choose the engine and run the decompile script. The script handles APK, XAPK, AAB, DEX, JAR, and AAR files.

bash ${CLAUDE_PLUGIN_ROOT}/skills/android-reverse-engineering/scripts/decompile.sh [OPTIONS] <file>

For XAPK files (ZIP bundles containing multiple APKs, used by APKPure and similar stores): the script automatically extracts the archive, identifies all APK files inside (base + split APKs), and decompiles each one into a separate subdirectory. The XAPK manifest is copied to the output for reference.

For AAB files (Android App Bundles): the script uses bundletool to generate a universal APK from the bundle, then decompiles it. bundletool must be installed (run install-dep.sh bundletool).

For DEX files: jadx handles them natively. For Fernflower, dex2jar is used as an intermediate step (same as APK files).

Options:

• -o <dir> — Custom output directory (default: <filename>-decompiled)
• --deobf — Enable deobfuscation (recommended for obfuscated apps)
• --no-res — Skip resources, decompile code only (faster)
• --engine ENGINE — jadx (default), fernflower, or both

Engine selection strategy:

When using --engine both, the outputs go into <output>/jadx/ and <output>/fernflower/ respectively, with a comparison summary at the end showing file counts and jadx warning counts. Review classes with jadx warnings in the Fernflower output for better code.

For APK files with Fernflower, the script automatically uses dex2jar as an intermediate step. dex2jar must be installed for this to work.

See ${CLAUDE_PLUGIN_ROOT}/skills/android-reverse-engineering/references/jadx-usage.md and ${CLAUDE_PLUGIN_ROOT}/skills/android-reverse-engineering/references/fernflower-usage.md for the full CLI references.

Phase 3: Analyze Structure

Navigate the decompiled output to understand the app's architecture.

Actions:

1. Read AndroidManifest.xml from <output>/resources/AndroidManifest.xml:

   • Identify the main launcher Activity
   • List all Activities, Services, BroadcastReceivers, ContentProviders
   • Note permissions (especially INTERNET, ACCESS_NETWORK_STATE)
   • Find the application class (android:name on <application>)

2. Survey the package structure under <output>/sources/:

   • Identify the main app package and sub-packages
   • Distinguish app code from third-party libraries
   • Look for packages named api, network, data, repository, service, retrofit, http — these are where API calls live

3. Identify the architecture pattern:

   • MVP: look for Presenter classes
   • MVVM: look for ViewModel classes and LiveData/StateFlow
   • Clean Architecture: look for domain, data, presentation packages
   • This informs where to look for network calls in the next phases

Phase 4: Trace Call Flows

Follow execution paths from user-facing entry points down to network calls.

Actions:

1. Start from entry points: Read the main Activity or Application class identified in Phase 3.

2. Follow the initialization chain: Application.onCreate() often sets up the HTTP client, base URL, and DI framework. Read this first.

3. Trace user actions: From an Activity, follow:

   • onCreate() → view setup → click listeners
   • Click handler → ViewModel/Presenter method
   • ViewModel → Repository → API service interface
   • API service → actual HTTP call

4. Map DI bindings (if Dagger/Hilt is used): Find @Module classes to understand which implementations are provided for which interfaces.

5. Handle obfuscated code: When class names are mangled, use string literals and library API calls as anchors. Retrofit annotations and URL strings are never obfuscated.

See ${CLAUDE_PLUGIN_ROOT}/skills/android-reverse-engineering/references/call-flow-analysis.md for detailed techniques and grep commands.

Phase 5: Extract and Document APIs

Find all API endpoints and produce structured documentation.

Action: Run the API search script for a broad sweep.

bash ${CLAUDE_PLUGIN_ROOT}/skills/android-reverse-engineering/scripts/find-api-calls.sh <output>/sources/

Additional options:

• --context N — Show N lines of context around each match (recommended: --context 3)
• --report FILE — Export results as a structured Markdown report
• --dedup — Deduplicate results by endpoint/URL

Targeted searches:

# Only Retrofit
bash ${CLAUDE_PLUGIN_ROOT}/skills/android-reverse-engineering/scripts/find-api-calls.sh <output>/sources/ --retrofit

# Only hardcoded URLs
bash ${CLAUDE_PLUGIN_ROOT}/skills/android-reverse-engineering/scripts/find-api-calls.sh <output>/sources/ --urls

# Only auth patterns
bash ${CLAUDE_PLUGIN_ROOT}/skills/android-reverse-engineering/scripts/find-api-calls.sh <output>/sources/ --auth

# Only Kotlin coroutines/Flow patterns
bash ${CLAUDE_PLUGIN_ROOT}/skills/android-reverse-engineering/scripts/find-api-calls.sh <output>/sources/ --kotlin

# Only RxJava patterns
bash ${CLAUDE_PLUGIN_ROOT}/skills/android-reverse-engineering/scripts/find-api-calls.sh <output>/sources/ --rxjava

# Only GraphQL patterns
bash ${CLAUDE_PLUGIN_ROOT}/skills/android-reverse-engineering/scripts/find-api-calls.sh <output>/sources/ --graphql

# Only WebSocket patterns
bash ${CLAUDE_PLUGIN_ROOT}/skills/android-reverse-engineering/scripts/find-api-calls.sh <output>/sources/ --websocket

# Only security patterns (cert pinning, exposed secrets, debug flags)
bash ${CLAUDE_PLUGIN_ROOT}/skills/android-reverse-engineering/scripts/find-api-calls.sh <output>/sources/ --security

# Full analysis with context and Markdown report
bash ${CLAUDE_PLUGIN_ROOT}/skills/android-reverse-engineering/scripts/find-api-calls.sh <output>/sources/ --context 3 --dedup --report report.md

Then, for each discovered endpoint, read the surrounding source code to extract:

• HTTP method and path
• Base URL
• Path parameters, query parameters, request body
• Headers (especially authentication)
• Response type
• Where it's called from (the call chain from Phase 4)

Document each endpoint using this format:

### `METHOD /path`

- **Source**: `com.example.api.ApiService` (ApiService.java:42)
- **Base URL**: `https://api.example.com/v1`
- **Path params**: `id` (String)
- **Query params**: `page` (int), `limit` (int)
- **Headers**: `Authorization: Bearer <token>`
- **Request body**: `{ "email": "string", "password": "string" }`
- **Response**: `ApiResponse<User>`
- **Called from**: `LoginActivity → LoginViewModel → UserRepository → ApiService`

See ${CLAUDE_PLUGIN_ROOT}/skills/android-reverse-engineering/references/api-extraction-patterns.md for library-specific search patterns and the full documentation template.

Phase 6: Security Patterns Scan

Run the automated security sweep over the decompiled source to surface security-relevant patterns. This is the static baseline that Phase 8 (Security Analysis & Vulnerability Audit) later synthesizes together with the runtime findings from Phase 7.

Action: Run the security-focused search:

bash ${CLAUDE_PLUGIN_ROOT}/skills/android-reverse-engineering/scripts/find-api-calls.sh <output>/sources/ --security --context 3

Look for and flag:

• Disabled certificate pinning — custom TrustManager that trusts all certs, ALLOW_ALL_HOSTNAME_VERIFIER
• Exposed secrets — hardcoded passwords, API keys, encryption keys in source code
• Debug flags left on — BuildConfig.DEBUG checks, staging URLs, verbose logging
• Weak crypto — MD5 hashing, ECB mode encryption, hardcoded IVs/salts
• Network Security Config — check res/xml/network_security_config.xml for cleartextTrafficPermitted="true" or overly broad trust anchors

These raw findings feed Phase 8. Targeted vulnerability hunting (e.g. Fragment Injection) and the RASP/anti-tamper synthesis from Phase 7's dynamic analysis happen in Phase 8.

Phase 7: Dynamic Analysis with Frida (Adaptive Loop)

Use Frida to observe and modify app behavior at runtime. Do not use pre-built generic bypass scripts. Instead, generate custom Frida scripts based on what the static analysis (Phases 3–6) revealed in the decompiled code, then iterate based on crash logs and runtime behavior.

This phase requires a connected device/emulator. The user likely already has frida-server on their device — detect it first before offering to install anything.

Step 7.1: Setup Frida Environment

Action: Run the Frida setup script. It detects the existing environment before changing anything.

bash ${CLAUDE_PLUGIN_ROOT}/skills/android-reverse-engineering/scripts/setup-frida.sh

The script performs these checks in order:

1. adb connectivity — is a device/emulator connected?
2. frida-server on device — checks common paths (/data/local/tmp/frida-server, etc.) and running processes
3. frida-server version — extracts version from the binary on device
4. Python 3 + venv module — required for frida-tools
5. Creates/reuses venv — at ~/.local/share/frida-re/venv, installs frida-tools matching the device's frida-server version
6. Version match validation — warns if client/server versions diverge
7. Connectivity test — runs frida-ps -U to verify end-to-end

If frida-server is missing from the device, the script prints instructions. To auto-install:

bash ${CLAUDE_PLUGIN_ROOT}/skills/android-reverse-engineering/scripts/setup-frida.sh --install-server

Important: The venv ensures frida-tools never pollutes the global Python environment. The version matching ensures client and server are compatible. If the user already has a working frida-server, the script adapts to their version instead of forcing an upgrade.

Read the machine-readable output lines (FRIDA_VENV=, FRIDA_SERVER_VERSION=, FRIDA_DEVICE=, FRIDA_STATUS=) to configure subsequent steps.

Step 7.2: Baseline Crash Check (Before Any Hooks)

Before writing any Frida script, check if the app even runs on this device. Many apps with RASP will crash immediately on rooted devices/emulators.

Action: Launch the app and capture crash diagnostics.

bash ${CLAUDE_PLUGIN_ROOT}/skills/android-reverse-engineering/scripts/adb-crash-capture.sh -p <package>

Options:

• -t <seconds> — monitoring window (default: 10)
• -a <activity> — launch specific activity instead of auto-detect
• -o <dir> — save logcat/crash logs to directory
• -v — include full logcat in output

Read the machine-readable output:

• APP_STATUS=running — app is fine, proceed to runtime analysis
• APP_STATUS=crashed — check CRASH_SIGNAL, CRASH_EXCEPTION, CRASH_MESSAGE
• APP_STATUS=exited — app quit without a visible crash (common RASP pattern: System.exit() or Process.killProcess())

The script also outputs:

• JAVA CRASH section — full stack trace from FATAL EXCEPTION
• NATIVE CRASH section — signal info and native backtrace
• RASP/SECURITY INDICATORS — log lines mentioning security, root, frida, tamper, integrity, debug, hook
• APP LOG — last 50 lines from the app's PID

If the app runs fine (status=running), skip to Step 7.4 for runtime analysis. If the app crashes or exits, proceed to Step 7.3.

Step 7.3: Adaptive Bypass Loop

This is the core of dynamic analysis. Use the decompiled code from previous phases combined with crash logs to understand WHY the app is dying, then generate a targeted Frida script to bypass that specific check.

The loop:

1. Read crash output from Step 7.2 (or previous iteration)
2. Identify the protection mechanism:
   - Cross-reference crash class/method with decompiled code
   - Follow the stack trace back to the triggering check
   - Look for the RASP/security indicators in logs
3. Read the relevant decompiled source to understand the check logic
4. Generate a Frida script that specifically disables that check
5. Run the script and capture new crash output
6. If still crashing: repeat from step 1 (new crash = new check to bypass)
7. If running: proceed to Step 7.4

How to identify protection mechanisms from crash data:

How to generate the bypass script:

Read the decompiled source of the method that performs the check. Understand:

• What does the method return? (boolean, int, void)
• Is it Java or native?
• Does it run once (in onCreate) or continuously (background thread)?
• What is the expected "safe" return value?

Then write a Frida script that hooks that specific method. Examples of patterns (adapt to actual code):

For a Java method returning boolean:

Java.perform(function() {
    var cls = Java.use('com.example.security.RootChecker');
    cls.isDeviceRooted.implementation = function() {
        console.log('[bypass] RootChecker.isDeviceRooted() called, returning false');
        return false;
    };
});

For a native function:

var funcAddr = Module.findExportByName("libsecurity.so", "Java_com_example_NativeCheck_verify");
if (funcAddr) {
    Interceptor.replace(funcAddr, new NativeCallback(function() {
        console.log('[bypass] native verify() called, returning 0');
        return 0;
    }, 'int', []));
}

For System.exit() prevention (when you don't know the exact check yet):

Java.perform(function() {
    var System = Java.use('java.lang.System');
    System.exit.implementation = function(code) {
        console.log('[bypass] System.exit(' + code + ') blocked');
        // Don't call the original — app stays alive
        // Check the stack trace to find who's calling this:
        console.log(Java.use("android.util.Log").getStackTraceString(
            Java.use("java.lang.Throwable").$new()
        ));
    };
});

Action: Save the generated script to a temp file and run it:

bash ${CLAUDE_PLUGIN_ROOT}/skills/android-reverse-engineering/scripts/frida-run.sh \
  -p <package> -l /tmp/bypass.js -t 15

Options:

• -t <seconds> — timeout (default: 30)
• --pause — suspend app on spawn, useful for early hooks before any app code runs
• --attach — attach to running process instead of spawning
• -e "<code>" — inline JavaScript instead of a file
• --output-dir <dir> — save stdout/stderr/crash to files

Read the output:

• FRIDA_RESULT=success — bypass worked, app is running
• FRIDA_RESULT=crash — check the CRASH LOG section, identify the next check to bypass
• FRIDA_RESULT=connection_failed — frida-server issue, re-run setup-frida.sh
• FRIDA_RESULT=timeout — app ran for the full timeout window (usually means success)

If the app crashes again: read the new crash log, cross-reference with decompiled code, identify the NEXT protection check, add another hook to the script, and run again. Each iteration should bypass one more check. Common pattern: apps have 3–5 layered checks.

Important considerations:

• Use --pause (spawn gating) when the check runs in Application.onCreate() — this ensures hooks are in place before any app code executes
• If the app uses a background thread for continuous checks, hook Thread.start() or the specific Runnable to neutralize it
• If native checks use ptrace(PTRACE_TRACEME), hook ptrace via Interceptor.replace
• If the app checks for frida-server's default port (27042), the frida-server can be started on a different port: frida-server -l 0.0.0.0:1337
• Log stack traces in your hooks — they reveal the call chain leading to the check, which helps find related checks

Step 7.4: Runtime Analysis

Once the app is running (with or without bypass scripts), use Frida for the actual analysis goals.

Generate analysis scripts based on what static analysis found. Do not use generic scripts — target the specific classes, methods, and patterns identified in Phases 3–6.

Common analysis patterns (adapt to the actual code found):

Intercept HTTP traffic — hook the specific HTTP client the app uses (identified in Phase 5):

// Example: if static analysis found OkHttp usage
Java.perform(function() {
    var OkHttpClient = Java.use('okhttp3.OkHttpClient');
    var RealCall = Java.use('okhttp3.internal.connection.RealCall');
    // Hook based on actual classes found in the decompiled code
});

Monitor crypto operations — hook the specific encryption methods found in Phase 6:

// Example: if static analysis found AES usage in com.example.crypto.CryptoHelper
Java.perform(function() {
    var helper = Java.use('com.example.crypto.CryptoHelper');
    // Hook the specific encrypt/decrypt methods found
});

Trace method calls — when static analysis shows a call flow but you need to verify it at runtime:

# Use frida-trace for quick method tracing (uses the venv)
$FRIDA_VENV/bin/frida-trace -U -f <package> -j 'com.example.api.*!*'

Action: Write the analysis script, run it, and interpret results:

bash ${CLAUDE_PLUGIN_ROOT}/skills/android-reverse-engineering/scripts/frida-run.sh \
  -p <package> -l /tmp/analysis.js -t 60 --output-dir ./frida-output/

If bypass hooks are needed alongside analysis hooks, combine them in a single script — bypass hooks first, then analysis hooks.

Step 7.5: Iterate and Document

After each Frida run, document:

• What protection was found and how it was bypassed
• What runtime behavior was observed that static analysis couldn't show
• Any new endpoints, keys, or tokens discovered at runtime

Feed runtime findings back into the API documentation from Phase 5 — runtime analysis often reveals:

• Dynamically constructed URLs that don't appear in static code
• Encryption keys loaded from server responses
• Token refresh flows that are only triggered under specific conditions
• Feature flags that change API behavior

Phase 8: Security Analysis & Vulnerability Audit

Synthesize the automated sweep from Phase 6 with the runtime findings from Phase 7, then run targeted vulnerability detectors for issues the generic sweep does not catch.

Carry forward from Phase 6: review every flag the find-api-calls.sh --security sweep produced (disabled certificate pinning, exposed secrets, debug flags, weak crypto, Network Security Config) and confirm each by reading the surrounding decompiled code — raw grep hits are leads, not findings, until you read the code.

Carry forward from Phase 7:

• RASP/Anti-tamper — document what protections were found, how robust they are, and what was needed to bypass them

Targeted vulnerability hunting: run the issue-specific detectors that the generic sweep does not cover. Each has its own playbook in references/.

Fragment Injection (exported PreferenceActivity)

Android PreferenceActivity honours an Intent extra (:android:show_fragment) that selects which Fragment class to instantiate. An exported (or launchable) activity that does this without a restrictive isValidFragment() override lets an attacker instantiate arbitrary internal fragments inside the victim's process/UID — bypassing UI gating and access controls.

Action: Run the detector against the decompiled output.

bash ${CLAUDE_PLUGIN_ROOT}/skills/android-reverse-engineering/scripts/find-fragment-injection.sh \
  <output>/ --report <output>/fragment-injection-report.md --json <output>/fragment-injection.json

Read the machine-readable output:

• TARGET_SDK_VERSION=<n> — drives the isValidFragment gate (see below).
• FRAG_INJECTION_CANDIDATE=<class> — an exported/launchable PreferenceActivity that is injectable: always_true override, OR missing override with targetSdk < 19 (or unknown). Exit 0 if any candidate exists, exit 2 if none.
• FRAG_INJECTION_BROKEN=<class> — missing override with targetSdk ≥ 19: the app crashes on injection instead of being exploited. Broken, not vulnerable — report as a robustness issue.
• IS_VALID_FRAGMENT=<class>:<status> — missing, always_true, or whitelist (only whitelist is safe).
• SHOW_FRAGMENT_READER / DYNAMIC_FRAGMENT_LOAD — code that reads the extra or does Fragment.instantiate/FragmentFactory/Class.forName from Intent extras. DYNAMIC_FRAGMENT_LOAD is the relevant surface for modern AndroidX apps (which don't use PreferenceActivity); ANDROIDX_PREFERENCE flags when the modern preference API is in use.

For each candidate, confirm dynamically: launch the activity via adb with the :android:show_fragment extra set to an internal fragment, and observe logcat/Frida for the fragment instantiating (lifecycle logs or a fragment-stack crash). Use --pause Frida spawn gating to hook the candidate subclass's isValidFragment/Fragment.instantiate before onCreate.

See ${CLAUDE_PLUGIN_ROOT}/skills/android-reverse-engineering/references/android-fragment-injection.md for the full detection-→-exploitation-→-classification playbook, adb and Frida commands, and remediation guidance.

Phase 9: Firebase & Google API Testing (Conditional)

Only run this phase if the decompiled app contains Google API keys or Firebase configuration. If none are present, skip it entirely — do not invent keys, do not hit Google endpoints speculatively.

This phase is strictly for apps the user is authorized to test (their own apps, signed engagements, or bug-bounty programs that explicitly permit it). Confirm authorization before running write/create probes (Realtime DB PUT, Dynamic Links creation, FCM send) or billable probes (Maps, Vision, Translate, etc.).

Step 9.1: Detect Firebase/Google configuration (gatekeeper)

Action: Run the detection script against the decompiled output. Its exit code drives the rest of the phase.

bash ${CLAUDE_PLUGIN_ROOT}/skills/android-reverse-engineering/scripts/find-firebase-config.sh \
  <output>/ --env /tmp/fb-env.sh --json /tmp/fb-env.json

The script scans res/values/strings.xml, AndroidManifest.xml, assets/google-services.json, assets/appsettings.json, and the full decompiled tree for values matching AIza[0-9A-Za-z_\-]{35} and known Firebase keys (google_api_key, project_id, firebase_database_url, gcm_defaultSenderId, google_app_id, google_storage_bucket, default_web_client_id).

Read the machine-readable output:

• FIREBASE_FOUND=true or GOOGLE_API_KEY_FOUND=true (exit 0) → proceed to Step 9.2.
• Both false (exit 2) → skip the rest of Phase 9 entirely and move on to the final deliverables.

The --env file produced on success is ready to source before running any Firebase/Google request (API_KEY, PROJECT_ID, DB_URL, APP_ID, GCM_SENDER_ID, PACKAGE, OAUTH_CLIENT_ID, STORAGE_BUCKET, plus an API_KEYS array for apps that ship multiple keys).

Step 9.2: Confirm authorization

Before running any probe, confirm with the user that the app is in scope for Firebase/Google API testing. If the user has not authorized this, stop and report the configuration findings only. Do not proceed to Step 9.3.

Step 9.3: Run the test matrix

Action: Run the automated matrix against the extracted configuration.

bash ${CLAUDE_PLUGIN_ROOT}/skills/android-reverse-engineering/scripts/test-firebase-google.sh \
  --env /tmp/fb-env.sh --report <output>/firebase-google-report.md

Useful flags:

• --skip-billable — skip Section 9 (Maps/AI/YouTube) to avoid billable calls on the target project.
• --skip-writes — skip Realtime DB PUT, Dynamic Links creation, and FCM send (use when authorization covers reads only).
• --only auth,rtdb,firestore — restrict to specific sections (auth, rtdb, firestore, remoteconfig, storage, dynamiclinks, fcm, gemini, billable).
• --api-key <KEY> — override the key (useful when iterating through multiple keys from API_KEYS).

The script runs the full playbook in ${CLAUDE_PLUGIN_ROOT}/skills/android-reverse-engineering/references/firebase-google-api-testing.md — Firebase Auth (signup, signin, OIDC providers, phone, OOB codes, enumeration), Realtime DB (unauth and authenticated reads, rules, write test), Firestore (list root, common collections, authenticated Bearer), Remote Config (firebase:fetch), Cloud Storage (both bucket conventions), Dynamic Links (open redirect / phishing check), FCM legacy send, Gemini (/files, /models, /cachedContents, gemini-pro:generateContent — TruffleSecurity vector), and billable Maps/AI/YouTube/Cloud Functions probes.

Each probe is classified as VULNERABLE, SAFE, BLOCKED, NOT_FOUND, OK, ERROR-200, INFO, or NETWORK_ERROR. When Firebase Auth returns an idToken (anonymous or email signup open), the script captures it and reuses it as $JWT for the authenticated Realtime DB / Firestore / lookup probes in the same run.

Read the final machine-readable output:

• PROBE_COUNT=<n> — how many endpoints were exercised.
• VULN_COUNT=<n> — how many returned exploitable data.
• REPORT_FILE=<path> — Markdown report with per-probe status, verdict, and response excerpt.

Step 9.4: Iterate across multiple keys

If find-firebase-config.sh reported API_KEY_COUNT > 1, re-run test-firebase-google.sh with each additional key (source /tmp/fb-env.sh && test-firebase-google.sh --api-key "${API_KEYS[1]}" --report <output>/firebase-google-report-key2.md). Different keys often belong to different GCP projects with different APIs enabled — a key that looks safe on project A may be wide open on project B.

Interpreting results

See ${CLAUDE_PLUGIN_ROOT}/skills/android-reverse-engineering/references/firebase-google-api-testing.md for the full response-interpretation table (SERVICE_DISABLED, PERMISSION_DENIED, ADMIN_ONLY_OPERATION, OPERATION_NOT_ALLOWED, NO_TEMPLATE, etc.) and for the raw curl commands to re-run any single probe manually.

At the end of the workflow, deliver:

1. Decompiled source in the output directory
2. Architecture summary — app structure, main packages, pattern used
3. API documentation — all discovered endpoints in the format above
4. Call flow map — key paths from UI to network (especially authentication and main features)
5. Security findings — certificate pinning status, exposed secrets, debug flags, crypto issues, Fragment Injection exposure (candidates, isValidFragment status, adb/Frida confirmation)
6. Dynamic analysis results (if Phase 7 was performed):
   • Protection mechanisms found and bypass scripts generated
   • Runtime-only discoveries (dynamic URLs, keys, tokens, feature flags)
   • Frida scripts used (saved in the output directory for reproducibility)
7. Firebase & Google API findings (if Phase 9 was performed):
   • Extracted configuration values (find-firebase-config.sh output / env file)
   • Per-probe verdicts from test-firebase-google.sh (the Markdown report)
   • Highlighted VULNERABLE findings with the response excerpt and impact
   • Any additional keys tested and their separate reports

Use --report report.md on find-api-calls.sh to generate a structured Markdown report automatically.

References

• ${CLAUDE_PLUGIN_ROOT}/skills/android-reverse-engineering/references/setup-guide.md — Installing Java, jadx, Fernflower/Vineflower, dex2jar, and optional tools
• ${CLAUDE_PLUGIN_ROOT}/skills/android-reverse-engineering/references/jadx-usage.md — jadx CLI options and workflows
• ${CLAUDE_PLUGIN_ROOT}/skills/android-reverse-engineering/references/fernflower-usage.md — Fernflower/Vineflower CLI options, when to use, APK workflow
• ${CLAUDE_PLUGIN_ROOT}/skills/android-reverse-engineering/references/api-extraction-patterns.md — Library-specific search patterns and documentation template
• ${CLAUDE_PLUGIN_ROOT}/skills/android-reverse-engineering/references/call-flow-analysis.md — Techniques for tracing call flows in decompiled code
• ${CLAUDE_PLUGIN_ROOT}/skills/android-reverse-engineering/references/firebase-google-api-testing.md — Phase 9 playbook: Firebase Auth, Realtime DB, Firestore, Remote Config, Storage, Dynamic Links, FCM, Gemini, billable Maps/AI probes
• ${CLAUDE_PLUGIN_ROOT}/skills/android-reverse-engineering/references/android-fragment-injection.md — Phase 8 playbook: detecting and exploiting Fragment Injection via exported PreferenceActivity, adb/Frida confirmation, result classification, remediation
• ${CLAUDE_PLUGIN_ROOT}/skills/android-reverse-engineering/references/setup-guide.md — Frida setup section covers Python venv, frida-server, and version matching