More screenshots

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIDeepSeekDeepSeekMuseMuse+22 more

    APK Reverse Engineering: CLI-First Triage to Patch

    1

    Reverse-engineer Android APKs from the command line: manifest triage, jadx and smali static analysis, Frida dynamic hooks, native .so detours, and surgical patch-and-rebuild.

    $5

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    APK Reverse Engineering: CLI-First Triage to Patch

    APK Reverse Engineering: CLI-First Triage to Patch

    Example session with this skill installed

    Does this fitness app send location data to a third party, and if so, when? The APK is on my machine; I own the app and want to understand its behavior before release.

    • Read your context and instructions
    • Compiled the apk reverse engineering

    FINDING — yes, it sends location, but only while in-app tracking is enabled, and the toggle works.

    Phase 1 — triage: manifest shows ACCESS_FINE_LOCATION, an exported receiver for boot completion, and a native library lib/arm64-v8a/libtracker.so. Signing certificate is the publisher's. The permission plus the native library make this a real question, not a fishing expedition.

    Phase 2 — static: string search for the analytics domain finds it in strings.xml, referenced from a class called TelemetryManager. jadx shows TelemetryManager collecting latitude/longitude and posting them in a JSON body. The call site sits inside the location callback, gated on a boolean trackingEnabled. Smali confirms the gate is a simple if-eqz branch — the decompilation is trustworthy here.

    Phase 3 — dynamic: Frida hook on the HTTP layer logs the request body. App started, location granted, two minutes of walking: three POSTs, each carrying lat/long with a timestamp. Then the in-app tracking setting toggled off: the POSTs stop. Hypothesis confirmed by runtime evidence — the app sends location only while tracking is enabled, and the toggle works.

    Phase 4 — not needed. The question was "what does it do," not "change it." No patch. The deliverable is the finding with its evidence: location is sent to the analytics endpoint while tracking is on, with the code location, the endpoint, the trigger, and the Frida logs attached.

    If the question had been a patch — forcing the toggle off by default — the smali edit would be a single branch change in the trackingEnabled initialization, one rebuild, one reinstall, and a re-run of the Frida hook to confirm the POSTs no longer fire.

    Connects securely to your tools. The creator never sees your data.

    About this skill

    Work from the question, not the tool. Every APK task is one of four things: what does it do (triage), how does this feature work (targeted static analysis), what happens at runtime (dynamic), or change the behavior (patch and rebuild). This skill picks the cheapest tool that answers the current question.

    Phase 1 is triage in minutes: unzip the contents, verify the signing certificate, decode with apktool, and read the manifest first — permissions and exported components tell you what the app can do, and native libraries decide whether the dynamic phase needs a native detour. Phase 2 is static analysis with jadx, starting from entry points and string searches rather than browsing the tree, and dropping to smali — the ground truth — whenever the decompiler's reconstruction looks wrong. Phase 3 is dynamic confirmation with Frida: hooks that log arguments and return values before anything is changed, plus the native detour through JNI entry points when the interesting logic lives in a .so. Phase 4 is patch and rebuild: one surgical smali edit per rebuild, sign, align, install, and re-test.

    The skill includes a triage checklist, a static findings template, a hook script template, a tool decision table, and a pre-delivery checklist. The standard is agreement: decompiler output, smali, and runtime behavior must all agree before you conclude anything.

    Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Security.

    • Heavily obfuscated apps lose readable class names; the work then runs from strings, entry points, and runtime hooks.
    • jadx sometimes reconstructs a class badly — smali is the ground truth and the fallback.
    • Frida can't attach if the app detects root, Frida, or a debugger; that check must be identified first.
    • Split APKs (base + config splits) must be decoded together or code is missed.
    • A self-updating app can replace the APK analyzed; the version worked on must be recorded.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before buying a skill.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    Money-back guarantee

    Doesn't do what its page promises? You get your money back.

    Trust & safety

    Security scan

    Security scan passed 3 days ago

    Permissions

    Terminal / Shell
    Listed3 days ago

    What's inside

    Frequently Asked Questions

    The manifest. Permissions and exported components tell you what the app can do before you decompile anything — and the native library list decides whether you need the .so detour at all.

    Possibly — jadx output is a reconstruction. Drop to the smali in the decoded tree; it is the ground truth. If smali and runtime disagree with the decompiler, trust smali and runtime.

    Almost always register counts in smali. Revert, go smaller, one change per rebuild — two changes and a crash means you don't know which one broke it.

    New in Security