More screenshots

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    Firmware Pentest: Authorized Audits End to End

    1

    Audit IoT and embedded firmware end to end: acquisition, triage, extraction, automated analysis, emulation, dynamic testing, and exploitation — authorized testing only.

    $5

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    Firmware Pentest: Authorized Audits End to End

    Firmware Pentest: Authorized Audits End to End

    Example session with this skill installed

    Audit this MIPS home router firmware (vendor 2.1.4) for us. We have written permission and the vendor download. What do you find?

    • Read your context and instructions
    • Compiled the firmware pentest

    Home router firmware 2.1.4 — audit produced with the skill

    Authorization confirmed in writing before the first command; firmware version 2.1.4 and hash recorded. Triage: uImage + squashfs-xz, entropy normal — plaintext, not encrypted. Extraction ladder stopped at step one: binwalk -eM extracted cleanly, so the rest of the ladder was not needed (it runs in order, not all at once).

    Static analysis: EMBA flagged an outdated web server and a setuid helper. Manual review found the helper concatenating an unsanitized hostname into a system() call — the kind of bug a scanner flags as suspicious but only a human confirms. User-mode QEMU ran the helper; a crafted hostname yielded command execution in emulation, on the target architecture and endianness, with gadgets taken from the actual binary.

    Report: exact version and hash, reachable command injection in the helper with the crashing input as proof, and the outdated web server as a hardening note — its CVE list verified against the actual version string, paths not shown reachable, so it stays a note, not a vulnerability. The flash backup rule was recorded for any live-device follow-up: no hardware touched without a full backup first. Rubric score: 9/10 — one point held back because full-system emulation was not set up; user-mode covered the helper, and the full-system boot is recorded as a follow-up for testing the init sequence.

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

    About this skill

    A firmware audit is a pipeline with gates: acquire, triage, extract, analyze, emulate, test dynamically, exploit. This skill runs it in order, with a fallback at every stage. Triage fingerprints the image first (chip family, SDK lineage, plaintext vs compressed vs encrypted) so extraction never runs blind. Extraction follows a fallback ladder — binwalk, unblob, filesystem-specific tools, then the bootloader decryption route — because extraction failure does not prove encryption, and SDK-specific packing defeats individual tools regularly. Static analysis pairs an EMBA sweep with a manual credential hunt, binary hardening checks, and a version inventory where every flagged CVE is verified for version and reachable path before it becomes a finding.

    Emulation covers user-mode and full-system, with the three classic failures (NVRAM, interface names, /dev nodes) checked in order and the architecture matched exactly, including MIPS endianness. Dynamic testing walks the web UI by hand, triages crashes with gdb, and fuzzes with AFL++. A worked example audits a MIPS home router end to end — a reachable command injection in a setuid helper, demonstrated in emulation, with the outdated web server kept as a note. Authorization is the first input, not an afterthought.

    • Authorized testing only — your own devices, client engagements with written permission, CTFs, and public test targets.
    • Extraction failure does not prove encryption; the ladder must run in full before that conclusion.
    • A crash without demonstrated impact is reported as a crash, not a vulnerability.
    • Findings without a reachable path are notes, not vulnerabilities — the report says which is which.

    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 any payment.

    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

    Not necessarily. Run the whole ladder — binwalk, unblob, filesystem-specific tools — before concluding anything. SDK-specific packing defeats individual tools regularly, and declaring "encrypted" after one failed tool is a conclusion the evidence does not support.

    No. EMBA output is triage, not findings. Each flagged CVE needs two verifications: the version string is the one actually bundled, and the vulnerable code path is reachable. A CVE in unreachable code is a note, not a vulnerability.

    Because impact demonstrated in emulation is proof, and hardware is where mistakes brick devices. The flash backup comes before any live-device testing — a bricked device with no backup ends the engagement.

    New in Security