Reverse engineering binaries with ghidra
Skill meltedinhex/analyst-ai-pack/skills/reverse-engineering-binaries-with-ghidra
Uses Ghidra to disassemble and decompile a binary, navigate to key routines via imports and strings, annotate decompiled code, and run headless scripts to automate extraction of C2, crypto, and config. Activates for requests to reverse engineer with Ghidra, decompile a binary, or script Ghidra headless analysis.From its SKILL.md
npx -y skills add meltedinhex/analyst-ai-pack --skill reverse-engineering-binaries-with-ghidraAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 21 stars21 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its file declares
Copied from the file, not written here
The file declares its own license as Apache-2.0. That is the author’s claim about this one file, and it is not the same thing as the license GitHub reports for the repository, which is listed with the other numbers below.
SKILL.md
3.5 KB, 632 tokens by cl100k_base, as published. Nobody here has run it
Reverse Engineering Binaries with Ghidra
When to Use
- Static/dynamic triage has identified functionality that needs code-level understanding.
- You need to reverse a C2 protocol, decryption routine, or custom obfuscation.
- You want to extract hardcoded config (C2, keys, campaign IDs) from compiled code, or automate extraction across many samples via headless scripts.
Do not use Ghidra as a first step for unknown samples — triage statically and behaviorally first, and unpack before decompiling a packed binary.
Prerequisites
- Ghidra 11.x with JDK 17+.
- An isolated analysis host.
- Familiarity with x86/x64 (or the target arch) and OS API conventions.
- The sample unpacked (decompiling packed code wastes effort).
Safety & Handling
- Ghidra performs static analysis and does not run the sample, but still work in the lab.
- Treat any dumped/derived payloads as live samples.
Workflow
Step 1: Import and auto-analyze
Create a project, import the binary (Ghidra detects PE/ELF/Mach-O and arch), and run auto-analysis with default analyzers.
Step 2: Navigate to interesting code
Start from imports (Symbol Tree) and strings (Defined Strings). Cross-reference (X) from a
suspicious API or string to its callers to reach the relevant function quickly.
VirtualAlloc/WriteProcessMemory -> injection logic
InternetOpen/WinHttp* -> C2
CryptEncrypt / custom XOR loops -> crypto/config decode
Step 3: Annotate the decompiler output
Retype variables, rename functions, and add comments as you understand the code. Good naming turns the decompiler into readable pseudo-C and compounds across a session.
Step 4: Automate with headless scripts
For repeatable extraction, run Ghidra headless with a post-script. The bundled Python uses
analyzeHeadless to export functions/strings to JSON:
python scripts/analyst.py headless --ghidra /opt/ghidra --bin sample.bin --out out.json
Step 5: Extract the target artifact
Recover the C2 string, key, or config and validate it against dynamic observations.
Validation
- Renamed/retyped routines produce coherent pseudo-C consistent with observed behavior.
- Extracted config (C2/keys) matches what the sample uses at runtime.
- Headless export reproduces the interactive findings.
Pitfalls
- Decompiling packed/obfuscated code — unpack first or the output is meaningless.
- Trusting decompiler types blindly; verify against the disassembly for crypto/pointer logic.
- Ignoring TLS callbacks/constructors that run before the entry point.
References
- See
references/api-reference.mdfor the headless wrapper. - Ghidra documentation and API (linked in frontmatter).
What ships with it: 3 files
3.9 KB alongside SKILL.md, 1 of them executable
references/
- api-reference.md1.2 KB
scripts/
- analyst.pyruns2.4 KB
- LICENSE340 B