- Home
- Skills
- Code Quality & Review
- decompile-and-verify
Works with the AI Agents you already use
Decompile And Verify
Decompile compiled software with an AI agent and prove every recovered function against the original, plus an assessment skill that measures how recoverable a build is.
Free
decompile-and-verify
Example session with this skill installed
I have a stripped Linux binary 'auth_tool'. I need to recover the function at 0x4012a0 and prove the C code is correct. Use emulation for verification.
- Read your context and instructions
- Compiled the decompile-and-verify
- Generated the data export
I have completed the triage and verification for the requested function. The binary was identified as an x86-64 ELF (stripped). The function at 0x4012a0 has been recovered and verified as equivalent using differential emulation.
Decompilation and Verification Result
Target: auth_tool (sha256: 4f3a...b912)
Function: FUN_004012a0
Oracle: verify_emulated.py (Differential Emulation)
Status: equivalent-tested
// Verified candidate: attempt_2.c
// Logic: FNV-1a hash comparison for license key validation
#include <stdint.h>
uint32_t FUN_004012a0(const uint8_t *input, uint32_t len) {
uint32_t hash = 0x811c9dc5;
for (uint32_t i = 0; i < len; i++) {
hash ^= input[i];
hash *= 0x01000193;
}
return (hash % 97 == 42);
}
Verification Evidence
| Metric | Value |
|---|---|
| Attempts | 2 |
| Total Time | 145s |
| Inputs Tested | 256 (random, boundaries, null) |
| Verdict | MATCH (CANDIDATE_OUT == ORIGINAL_OUT) |
Next steps
- Check callers of
0x4012a0to confirm the input buffer source. - Run
ledger.py addto commit this verification towork/ledger.csv. - Use
rank_functions.pyto identify the next highest-priority function for recovery.
ledger.csv
CSV · data export
Example file from a real run - the skill writes it into your workspace.
Connects securely to your tools. The creator never sees your data.
About this skill
Trusting a decompiler at face value leads to broken ports and incorrect audits. This skill package automates the decompilation of native and bytecode binaries while using mathematical and empirical oracles to prove the recovered source matches the original logic.
What it does
- Automated Decompilation: Extracts readable source from ELF, PE, Mach-O, .NET, Java, APK, Python bytecode, and WebAssembly targets.
- Verification Oracles: Proves function accuracy using machine-code matching, differential emulation, or trace-replay tests.
- Triage and Analysis: Identifies file formats, entry points, embedded secrets, and packer signatures automatically.
- Progress Ledger: Tracks every recovery attempt, verification status, and time-on-task with cryptographic pinning to the binary.
- Exposure Assessment: Measures a build's resistance to reverse engineering and generates reports mapped to OWASP MASVS-RESILIENCE.
How it works
- Triage: Run
triage.pyto pin the binary's SHA256 and identify the target architecture. - Export: Use Ghidra or language-specific routes to dump initial decompiled drafts and function lists.
- Verify: Select an oracle (e.g.,
verify_emulated.pyorverify_match.py) to test recovered source against original execution. - Iterate: Refine code attempts based on oracle failures until the logic is equivalent.
- Report: Generate a final exposure report or verified source ledger for the audit or port.
Frameworks & tools
This skill requires Python 3.9+, Ghidra with JDK 21+, and the Unicorn emulation engine. It supports native compilers like GCC and Clang for matching oracles and standard bytecode tools for managed targets.
Why this beats prompting it yourself
Generic LLM prompts often hallucinate logic or ignore compiler optimizations that change program flow. This skill forces the agent to verify every claim against the ground truth of the binary, ensuring recovered source actually runs the same way as the original code.
Use cases
- Recovering lost source code from legacy binaries or retro ROMs.
- Auditing third-party libraries for hidden backdoors or embedded credentials.
- Porting compiled logic from one architecture or language to another with high confidence.
- Benchmarking software hardening and obfuscation techniques against automated recovery.
Known limitations
Static assessment cannot measure runtime anti-tampering or dynamic analysis protections. Large, heavily optimized modern builds may require manual toolchain identification for matching oracles to succeed.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Code Quality & Review.
Static analysis only. Does not measure dynamic anti-tampering. Requires Ghidra for native targets. Emulated oracles limited to x86 and x86-64.
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
Free to use
No account required to browse
Trust & safety
Security scan
Security scan passed today
- Free with an Agensi account
Windows only
- Accounts
GitHub
- Software
Python
Node.jsrea-agents@electron/asar
Git
Permissions
Allowed Hosts