reverse-engineer – skill Dữ liệu & Khoa học cho agent AI

> ⚠️ AUTHORIZED USE ONLY

> This skill is for educational purposes or authorized security assessments only.

> You must have explicit, written permission from the system owner before using this tool.

> Misuse of this tool is illegal and strictly prohibited.

> Mandatory confirmation gate

> Before running any command that probes, exploits, changes, persists on, extracts data from, or attempts credential access against a target:

> 1. Ask the user to state the exact target URL, IP, account, or resource.

> 2. Ask the user to confirm written authorization and the permitted scope.

> 3. Show the exact command(s) and explain their expected effect.

> 4. Wait for explicit confirmation in the current conversation.

>

> Without that confirmation, remain read-only and provide defensive guidance only. Prefer a sandbox, disposable VM, or controlled lab.

Common RE scripting environments

• IDAPython (IDA Pro scripting)
• Ghidra scripting (Java/Python via Jython)
• r2pipe (radare2 Python API)
• pwntools (CTF/exploitation toolkit)
• capstone (disassembly framework)
• keystone (assembly framework)
• unicorn (CPU emulator framework)
• angr (symbolic execution)
• Triton (dynamic binary analysis)

Use this skill when

• Working on common re scripting environments tasks or workflows
• Needing guidance, best practices, or checklists for common re scripting environments

Do not use this skill when

• The task is unrelated to common re scripting environments
• You need a different domain or tool outside this scope

Instructions

• Clarify goals, constraints, and required inputs.
• Apply relevant best practices and validate outcomes.
• Provide actionable steps and verification.
• The referenced implementation playbook is not bundled. Use the self-contained methodology below and inspected tool documentation; do not claim unavailable supporting content was reviewed.

Analysis Methodology

Phase 1: Reconnaissance

1. File identification: Determine file type, architecture, compiler

2. Metadata extraction: Strings, imports, exports, resources

3. Packer detection: Identify packers, protectors, obfuscators

4. Initial triage: Assess complexity, identify interesting regions

Phase 2: Static Analysis

1. Load into disassembler: Configure analysis options appropriately

2. Identify entry points: Main function, exported functions, callbacks

3. Map program structure: Functions, basic blocks, control flow

4. Annotate code: Rename functions, define structures, add comments

5. Cross-reference analysis: Track data and code references

Phase 3: Dynamic Analysis

1. Environment setup: Isolated VM, network monitoring, API hooks

2. Breakpoint strategy: Entry points, API calls, interesting addresses

3. Trace execution: Record program behavior, API calls, memory access

4. Input manipulation: Test different inputs, observe behavior changes

Phase 4: Documentation

1. Function documentation: Purpose, parameters, return values

2. Data structure documentation: Layouts, field meanings

3. Algorithm documentation: Pseudocode, flowcharts

4. Findings summary: Key discoveries, vulnerabilities, behaviors

Response Approach

When assisting with reverse engineering tasks:

1. Clarify scope: Ensure the analysis is for authorized purposes

2. Understand objectives: What specific information is needed?

3. Recommend tools: Suggest appropriate tools for the task

4. Provide methodology: Step-by-step analysis approach

5. Explain findings: Clear explanations with supporting evidence

6. Document patterns: Note interesting code patterns, techniques

Code Pattern Recognition

Common Patterns

// String obfuscation (XOR)
for (int i = 0; i < len; i++)
    str[i] ^= key;

// Anti-debugging (IsDebuggerPresent)
if (IsDebuggerPresent())
    exit(1);

// API hashing (common in malware)
hash = 0;
while (*name)
    hash = ror(hash, 13) + *name++;

// Stack string construction
char s[8];
*(DWORD*)s = 0x6C6C6548;  // "Hell"
*(DWORD*)(s+4) = 0x6F;     // "o\0"

Calling Conventions

• x86 cdecl: Args on stack, caller cleans
• x86 stdcall: Args on stack, callee cleans
• x64 Windows: RCX, RDX, R8, R9, then stack
• x64 System V: RDI, RSI, RDX, RCX, R8, R9, then stack
• ARM: R0-R3, then stack

Security & Ethics

Authorized Use Only

• Security research with proper authorization
• CTF competitions and educational challenges
• Malware analysis for defensive purposes
• Vulnerability disclosure through responsible channels
• Understanding software for interoperability

Never Assist With

• Unauthorized access to systems
• Creating malware for malicious purposes
• Bypassing software licensing illegitimately
• Intellectual property theft
• Any illegal activities

Example Interactions

CTF Binary Challenge

User: "I have a CTF binary that asks for a password. How do I approach this?"

Response: Let me guide you through the analysis:

1. First, identify the binary type:
   $ file challenge
   $ checksec --file=challenge

2. Load in Ghidra/IDA and find main()

3. Look for:
   - String comparisons (strcmp, memcmp)
   - Crypto functions (xor loops, constants like 0x5A827999)
   - Input validation logic

4. Set breakpoints on comparison functions in GDB:
   $ gdb ./challenge
   (gdb) break strcmp
   (gdb) run

5. Examine memory at comparison point to see expected value

Library Analysis

User: "I need to understand how this closed-source DLL handles authentication"

Response: For authorized analysis of the DLL:

1. Static analysis:
   - List exports: dumpbin /exports library.dll
   - Find auth-related functions by name patterns
   - Load in IDA/Ghidra, analyze exported functions

2. Dynamic analysis:
   - Hook API calls with Frida
   - Monitor network traffic
   - Trace function parameters

3. Documentation:
   - Document function signatures
   - Map data structures
   - Note any security considerations

Limitations

• Use this skill only when the task clearly matches the scope described above.
• Do not treat the output as a substitute for environment-specific validation, testing, or expert review.
• Stop and ask for clarification if required inputs, permissions, safety boundaries, or success criteria are missing.