More screenshots

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    IDA Reverse Engineering Assistant

    1

    A disciplined IDA Pro investigation method: survey the binary first, let imports and strings pick your targets, annotate as you decode, report findings cited to addresses.

    $5

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    IDA Reverse Engineering Assistant

    IDA Reverse Engineering Assistant

    Example session with this skill installed

    Triage helper_update.exe (3.4 MB, Windows x64) from a support bundle — we're authorized to analyze it; the vendor is unresponsive and we need to know what it does before it stays on customer machines. Survey data so far: imports include WinHttpOpen/WinHttpSendRequest, CryptEncrypt, RegSetValueExW, CreateServiceW, IsDebuggerPresent, GetTickCount64. Strings include a Run-key registry path, 'telemetry', '/v1/report', and a base64-looking blob of about 400 characters.

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

    Survey verdict (before any deep read): a network-capable updater/agent with persistence options and mild anti-analysis. Two-sentence profile: it can install a lasting presence through registry or service mechanisms, communicates over HTTP, and encrypts something before sending. That is the map, not the territory — everything below is graded by the claim-strength rubric.

    Import buckets: networking (WinHttpOpen, WinHttpSendRequest); crypto (CryptEncrypt); persistence surface (RegSetValueExW, CreateServiceW); environment checks (IsDebuggerPresent plus GetTickCount64 — that pairing suggests timing-based debugger detection; level 1, rumor, until the code is read).

    Threads pulled, in order:

    1. Xrefs to the Run-key string → a single function, renamed install_persistence after decompilation (level 3, finding): it writes the executable path under the Run key only when a config flag byte is set. Persistence is real but conditional, and the report says exactly that, with the address.
    2. Xrefs to '/v1/report' → build_report_request: assembles a POST whose JSON body comes from a struct typed during annotation (machine ID, version, error counters). The 400-character blob flows into CryptEncrypt's caller; the key source is not yet decoded, so it is logged as a gap, not guessed at.
    3. CreateServiceW has no xrefs from reachable code — dead path or dynamically resolved. Recorded at level 1 and explicitly not reported as a capability.

    Report excerpt — Capabilities: conditional Run-key persistence (finding, install_persistence); HTTP reporting of machine and error telemetry (finding, build_report_request and its caller); payload encryption before send (indicated; key handling open). Gaps: the encryption key source, the unreachable service path, and whether the timing check gates behavior — each with the step that would close it. Recommendation attached to the evidence: the binary is a telemetry-reporting updater agent, not a silent implant; keep or remove it based on whether the vendor discloses the reporting, which your ticket can now ask about specifically.

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

    About this skill

    IDA rewards method and punishes wandering. This skill is the fixed investigation order that takes you from "unknown binary" to a defensible behavior profile. Step one is a 15-minute survey — architecture and format, entry point and pre-main init chain, the import table classified into capability buckets, a string harvest, exports — written down as evidence; if you can't summarize the binary's probable purpose in two sentences afterward, the survey was too shallow. Step two lets that evidence pick targets: xref the most suspicious string or import, decompile those functions first, and follow the chain instead of reading top-to-bottom. A decision table maps common survey signals (network imports plus URL strings, crypto plus opaque blobs, an almost-empty import table) to the right first move — including when the right move is to stop static reading and plan an unpack or dynamic pass. Step three is annotation discipline: rename everything you understand, comment at decision points with what the code means, type the structures, and tag evidence as you go so the report assembles itself. Step four covers the classic obstacles — packers, flattened control flow, encrypted strings, C++ library noise — plus automation hygiene for IDA's scripting layer: health-checked services, generous timeouts instead of assumed hangs, idempotent scripts, and never killing the GUI to fix a stuck worker. A claim-strength rubric (rumor / indicated / finding / confirmed) keeps the final report honest: every capability cites a function and an address, IOCs are extracted rather than guessed, and a gaps section names what would close each open question. Includes a complete worked example where the survey alone identified a credential-theft binary before any logic was read. For binaries you own or are authorized to analyze: incident response, interop, lost-source recovery, vendor verification.

    • Static-first method: behavior that only appears at runtime (dynamically resolved imports, environment-gated paths) is graded and flagged as a gap, not resolved — closing it needs a dynamic pass.
    • Packed samples must be unpacked before the survey means anything; the skill tells you when you're looking at a packer, but doesn't unpack for you.
    • Modern Go and Rust binaries leak runtime structure this general method doesn't exploit; a runtime-first recovery pass serves those better.
    • Findings are only as strong as their claim level — level-1 items are leads, not conclusions, and the report format keeps them labeled that way.

    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 24 days ago

    Permissions

    No special permissions declared or detected

    Listed24 days ago
    Updated21 days ago

    What's inside

    Frequently Asked Questions

    The investigation order, annotation discipline, and claim-strength rubric transfer to any disassembler with a decompiler. The automation-hygiene section is IDA-specific, but the survey → threads → annotate → report method is the substance of the skill.

    Budget. A real binary is mostly library code and logic you'll never need. The survey concentrates your reading on the roughly 5% of functions the evidence points at — in the worked example, four decoded functions produced the full behavior profile.

    The claim-strength rubric. Every capability is labeled rumor, indicated, finding, or confirmed based on the evidence behind it, and must cite a function name and address at finding level. Strings alone never get reported as behavior.

    New in Security