More screenshots
Works with the AI Agents you already use
Reverse Engineering: Compiled Target Analysis
Understand compiled, packed, or obfuscated programs the way you'd debug them: state a hypothesis, run the cheapest test that could kill it, and report only what the evidence supports.
$5
Reverse Engineering: Compiled Target Analysis
Example session with this skill installed
I'm working through a crackme from a training set — it takes a serial in the form XXXX-XXXX and just says valid or invalid. I don't want the answer handed to me; I want to understand how the check works so I can write it up. Where do I start?
- Read your context and instructions
- Compiled the reverse engineering
GUIDED ANALYSIS — crackme from a training set; built to be analyzed.
Triage first, because it sizes the problem. Native ELF, x86-64, not stripped, not packed (normal entropy, no packer sections). Strings give you serial:, valid, and invalid — and notably do NOT contain any stored serial. Imports: printf, scanf, memcmp, nothing else of interest. Conclusion before any disassembly: the expected value is computed, not stored — the most useful fact in the first fifteen minutes, learned from what's absent.
Two hypotheses, written down before testing. H1: the program transforms your input and compares the result against a fixed value. H2: it transforms a built-in value and compares against your raw input. They predict different things at the comparison — cheap to separate.
Finding the check: the invalid string's cross-reference lands in one function — call it check_serial. Inside it: a byte-loop call, then memcmp, then the branch that prints one message or the other. That's the parse-validate-act shape, and the memcmp is the final comparison, so the breakpoint goes there — not on the first compare you spot; challenge binaries plant decoys.
The cheapest test kills H2 in one run: breakpoint at the memcmp, enter AAAA-AAAA. At the break, one buffer holds your input after a transformation — each byte changed — and the other holds a fixed 9-byte array that never varies. Your input is what's transformed: H1 confirmed, H2 dead, and the transform is the byte-loop function you already walked past.
Now the static read is fast because it has a purpose: the loop XORs byte i with (0x5A + i) and rotates left by one. Direction check against the known pair from the breakpoint — the decoder must invert the fixed array: rotate right, then XOR. Result: K7QX-29MD. Verified as the skill requires — run the binary, enter it, get valid; that successful run is the second verification pair.
Write-up skeleton delivered with it: behavior (format gate, transform, memcmp) with evidence per claim, the H1/H2 chain and separating test, the recovered serial as a decoded artifact, and the one unknown left — banner code, unread and listed as unread.
Connects securely to your tools. The creator never sees your data.
About this skill
Reverse engineering fails in a predictable way: an analyst opens the disassembly with no question in mind and reads until exhaustion. This skill replaces that with the debugging loop — hypothesis, cheapest disproving test, update — applied to targets you own or are authorized to analyze: your own binaries, lab samples, CTF and crackme challenges, approved assessments. It opens with ground rules that are enforced throughout: authorization stated up front, work on copies with the original hashed and pristine, sandbox assumptions for unknown samples, and understanding as the deliverable — never weaponization. The triage pass decides in fifteen minutes whether a target is a ten-minute or ten-hour problem, using identity, strings, import surface, and an isolated smoke run. Entry-point finding follows the signposts: string cross-references, import call sites, the parse-validate-act main flow, and comparisons — with breakpoints placed at the final comparison because decoy checks are common. Before deep static work, the dynamic shortcut ladder offers five rungs, from letting the program compute the expected value and reading it at the comparison, through comparison hooking and syscall tracing, to symbolic execution and emulation. Transform decoding covers the recurring patterns — repeating-key XOR, single-byte XOR, standard crypto with a recoverable key, custom permutations — plus the direction check that prevents the classic backwards-decoder bug and a two-pair verification rule. Anti-analysis techniques are inventoried as findings in themselves. A worked example recovers a crackme's serial by killing one hypothesis with a single breakpoint. The report format is fixed: behavior with evidence, the hypothesis chain, decoded artifacts, and unknowns each paired with the experiment that would resolve them.
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.
- VM-protected and heavily packed targets can push the work from hours to days; the skill's triage identifies that class early, but the method doesn't make the protection cheap to defeat analytically.
- The dynamic ladder assumes you can run the target in an isolated lab; targets with hardware or server dependencies need stubs, and conclusions drawn against stubs are labeled conditional.
- It ends at verified understanding. Recovering a check's logic on a challenge or your own binary is in scope; defeating protections on third-party software you don't control is not something this skill covers.
How to use
- 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
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 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 24 days ago
Permissions
No special permissions declared or detected