More screenshots

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIDeepSeekDeepSeekMuseMuse+22 more

    Thick Client Security: Test the App You Ship to the Attacker

    1

    Security-test desktop applications from the outside in — map trust boundaries, attack local storage and IPC, intercept traffic, and report only reproducible findings.

    $5

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    Thick Client Security: Test the App You Ship to the Attacker

    Thick Client Security: Test the App You Ship to the Attacker

    Example session with this skill installed

    An Electron desktop app is about to ship. Assess it from the outside in and report only findings you can reproduce.

    • Read your context and instructions
    • Compiled the thick client security

    Authorization was confirmed in writing first — target, scope, and rules of engagement — before anything was touched. The trust boundary was mapped: main process, two helpers, and an updater; outbound domains listed; local state enumerated — config files, a SQLite cache, logs; and the privilege of each component noted. Local storage produced the first real finding: an API token in a plaintext config file, readable by any user on a shared machine, with the permissions and the exact path recorded and the token's scope demonstrated — it granted read access to the account. IPC produced the second: a named pipe that accepted commands from any local process with no authentication. A malformed input test showed it parsed without validation. The impact was demonstrated, not theorized: a local process could trigger a privileged action through it. Traffic interception showed the client used the system proxy with standard TLS, so visibility was straightforward — and the API behind it still enforced authorization server-side, so the hidden admin endpoints the client concealed were not reachable. That is a non-finding, recorded as such. The update channel checked out: updates are signed, and the signature is validated — not merely checked for presence — before install, with a tamper-evident channel. The Electron triage found the IPC bridge exposed in the preload script, which is where the pipe finding traced back to. The report lists two findings, each with the boundary crossed, exact reproduction steps, realized impact, and the minimal fix with a re-test procedure. No speculative hardening advice was padded in. The plaintext token and the unauthenticated pipe are the report; the things that held are recorded as held, not as findings.

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

    About this skill

    A desktop application is a server you ship to the attacker. Everything it trusts — local files, its own memory, the update channel, the hidden API endpoints — runs on hardware the adversary controls. Test it with that assumption, or test nothing. This skill starts with authorization in writing, then maps the trust boundary before testing anything: processes, network, local state, and privilege. It covers the local attack surface — secrets and config, local databases, binary planting, debug flags — the IPC channels scanners miss, intercepting traffic and questioning what the client trusts, and the update channel: the highest-value target in the client, because it can replace the application itself. It triages effort by tech stack — Electron, .NET, native — and ends with the reporting rule: only what you can reproduce, with the trust boundary crossed, the exact steps, the realized impact, and the minimal fix. Zero findings is a valid outcome; padding is not.

    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.

    Authorized assessments only — the target, scope, and rules of engagement go in writing before any work starts.
    A finding needs demonstrated impact, not theory; speculative hardening advice without a path is not a finding.
    Certificate pinning complicates interception; the pinning implementation itself becomes a test target, worked around with instrumentation.

    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

    No special permissions declared or detected

    Listed3 days ago

    What's inside

    Frequently Asked Questions

    Everything it trusts runs on hardware the adversary controls — local files, memory, the update channel. Testing must assume that, or it tests nothing.

    The updater. It can replace the application itself, so signed updates with real signature validation are the first thing to verify.

    A trust boundary crossed, exact reproduction steps, realized impact, and a minimal fix with a re-test procedure. Zero findings is a valid outcome; padding is not.

    New in Security