跳到主内容
智客 ZICQ

技能库 智客分类:Agent 工作流 rev-idapython

Rev Idapython

逆向工程的IDAPython和IDALib脚本参考. 当用户需要在IDA中写入IDAPython脚本,使用IDALib进行无头分析,在IDB数据库中操作,与IDA调试,操纵内存/注册,穿行函数/块/指令,与Hex-Rays解编解编译API合作,处理疏漏,或分批处理二进制.

1602 安装量

官方网址:skills.sh

技能介绍

先看中文介绍;官方 description 原文单独保留,不改写 SKILL.md。

做什么

逆向工程的IDAPython和IDALib脚本参考. 当用户需要在IDA中写入IDAPython脚本,使用IDALib进行无头分析,在IDB数据库中操作,与IDA调试,操纵内存/注册,穿行函数/块/指令,与Hex-Rays解编解编译API合作,处理疏漏,或分批处理二进制.

何时用

官方 description 未单独写出 Use when。按规范,代理会在用户任务与这段 description 的关键词匹配时激活本技能。

代理如何加载

按 Agent Skills 渐进披露:启动时只加载 name 与 description(约 100 token);任务匹配后才读入整份 SKILL.md 正文;scripts/、references/、assets/ 仅在需要时再读。 本文件正文结构:rev-idapython - IDAPython / IDALib Script Reference、Common API、Register Operations、Debug Memory Operations、Local Memory Operations (modifies IDB database)、Disassembly。

文件分析

文件分析:这是一份仅含 SKILL.md 的指令型技能,代理激活后整份正文进入上下文。 正文约 794 行,超过规范建议的 500 行,激活时仍会整份载入。

官方 description(原文)

IDAPython and IDALib script reference for reverse engineering. Activate when the user needs to write IDAPython scripts in IDA, use IDALib for headless analysis, operate on IDB databases, debug with IDA, manipulate memory/registers, traverse functions/blocks/instructions, work with Hex-Rays decompiler API, handle obfuscation, or batch-process binaries.

rev-idapython - IDAPython / IDALib Script ReferenceCommon APIRegister OperationsDebug Memory OperationsLocal Memory Operations (modifies IDB database)DisassemblyAddress ConversionFunction Operationsenumerate all functionsCode SnippetsByte Pattern Searchfind_bytes_list("90 90 90 90 90")

来源分类:skills.sh agent-skill

SKILL.md 与 Agent 调用

官方规范 ↗
name
rev-idapython
description
IDAPython and IDALib script reference for reverse engineering. Activate when the user needs to write IDAPython scripts in IDA, use IDALib for headless analysis, operate on IDB databases, debug with IDA, manipulate memory/registers, traverse functions/blocks/instructions, work with Hex-Rays decompiler API, handle obfuscation, or batch-process binaries.
  1. 发现技能客户端向 Agent 提供名称与描述目录。
  2. 匹配与调用用户指定或任务匹配后,载入 SKILL.md 指令。
  3. 按需加载按步骤读取参考文档、使用脚本与素材。

具体调用语法与可用工具以目标 Agent 客户端为准。 查看调用机制说明 ↗

安装这个技能

Skills CLI ↗

先选择目标 Agent 和安装范围,保留技能包的附属文件,安装后检查客户端能否发现该技能。

交给 Agent 安装

复制安装指令给支持 Agent Skills 的代理,确认其中的目标目录与客户端匹配。

把 Agent Skill「rev-idapython」安装到我的项目:SKILL.md 原文与官方 description 见 https://zicq.com/zh/skills/skl-f79b3c2f28e7b0d3-%E4%BC%8A%E8%BE%BE%E5%B9%B3%E9%80%9A%E4%BF%AE%E8%AE%A2%E7%89%88.html
请存为 .cursor/skills/rev-idapython/SKILL.md 或 .claude/skills/rev-idapython/SKILL.md,frontmatter 的 name 与 description 保持原样,不要改写。

GitHub 完整包 ↗

终端安装 · Skills CLI

需要 Node.js 与 npx。先查看仓库技能列表,确认实际名称。

npx skills add 'https://github.com/p4nda0s/reverse-skills' --list

npx skills add 'https://github.com/p4nda0s/reverse-skills' --skill 'rev-idapython'

CLI 会交互选择目标 Agent,默认安装到项目;用户级安装使用 -g。先通过查看命令核对仓库内容,再用 npx skills list 检查已安装技能。

阅读排版
--- name: rev-idapython description: IDAPython and IDALib script reference for reverse engineering. Activate when the user needs to write IDAPython scripts in IDA, use IDALib for headless analysis, operate on IDB databases, debug with IDA, manipulate memory/registers, traverse functions/blocks/instructions, work with Hex-Rays decompiler API, handle obfuscation, or batch-process binaries. --- # rev-idapython - IDAPython / IDALib Script Reference IDAPython script snippets for IDA interactive use and IDALib headless analysis. Use as reference when generating IDAPython code. - **IDAPython**: scripts run inside IDA GUI (Script Command, plugin, or IDC console) - **IDALib**: headless mode introduced in IDA 9.0 — run analysis scripts without opening the IDA GUI --- ## Common API ### Register Operations ```python idc.get_reg_value('rax') idaapi.set_reg_val("rax", 1234) ``` ### Debug Memory Operations ```python idc.read_dbg_byte(addr) idc.read_dbg_memory(addr, size) idc.read_dbg_dword(addr) idc.read_dbg_qword(addr) idc.patch_dbg_byte(addr, val) idc.add_bpt(0x409437) # add breakpoint idaapi.get_imagebase() # get image base address ``` ### Local Memory Operations (modifies IDB database) ```python idc.get_qword(addr) idc.patch_qword(addr, val) idc.patch_dword(addr, val) idc.patch_word(addr, val) idc.patch_byte(addr, val) idc.get_db_byte(addr) idc.get_bytes(addr, size) idaapi.get_dword(addr) idc.get_strlit_contents # read string literal ``` ### Disassembly ```python GetDisasm(addr) # get disassembly text idc.next_head(ea) # get next instruction address idc.create_insn(addr) # c, Make Code ida_bytes.create_strlit # create string, same as 'A' key ida_funcs.add_func(addr) # p, create function idc.del_items(addr) # U, undefine ``` ### Address Conversion ```python idc.get_name_ea(0, '_sub_6051') # get address by function name ``` ### Function Operations ```python ida_funcs.get_func(ea) # get function descriptor # enumerate all functions for func in idautils.Functions(): print("0x%x, %s" % (func, idc.get_func_name(func))) ``` --- ## Code Snippets ### Byte Pattern Search ```python import ida_bytes import ida_idaapi import ida_funcs import idc # find_bytes_list("90 90 90 90 90") # find_bytes_list("55 ??") # returns list of matching addresses def find_bytes_list(bytes_pattern): ea = -1 result = [] while True: ea = idc.find_bytes(bytes_pattern, ea + 1) if ea == ida_idaapi.BADADDR: break result.append(ea) return result ``` ### Appcall - Call Debuggee Functions ```python # test check_passwd(char *passwd) -> int passwd = ida_idd.Appcall.byref("MyFirstGuess") res = ida_idd.Appcall.check_passwd(passwd) if res.value == 0: print("Good passwd !") else: print("Bad passwd...") ``` ```python # Explicitly create the buffer as a byref object s_in = Appcall.byref("SomeEncryptedBuffer") # Buffers are always returned byref s_out = Appcall.buffer(" ", SizeOfBuffer) # Call the debuggee Appcall.decrypt_buffer(s_in, s_out, SizeOfBuffer) # Print the result print "decrypted=", s_out.value ``` ```python loadlib = Appcall.proto("kernel32_LoadLibraryA", "int __stdcall loadlib(const char *fn);") hmod = loadlib("dll_to_inject.dll") getlasterror = Appcall.proto("kernel32_GetLastError", "DWORD __stdcall GetLastError();") print "lasterror=", getlasterror() getcmdline = Appcall.proto("kernel32_GetCommandLineA", "const char *__stdcall getcmdline();") print "command line:", getcmdline() ``` ### Cross References ```python for ref in idautils.XrefsTo(ea): print(hex(ref.frm)) # shorthand [ref.frm for ref in idautils.XrefsTo(start_ea)] ``` ### Basic Block Traversal ```python fn = 0x4800 f_blocks = idaapi.FlowChart(idaapi.get_func(fn), flags=idaapi.FC_PREDS) for block in f_blocks: print(hex(block.start_ea)) ``` ```python # successor blocks for succ in block.succs(): print hex(succ.start_ea) # predecessor blocks for pred in block.preds(): print hex(pred.start_ea) ``` ### Debug Memory Read/Write ```python def patch_dbg_mem(addr, data): for i in range(len(data)): idc.patch_dbg_byte(addr + i, data[i]) def read_dbg_mem(addr, size): dd = [] for i in range(size): dd.append(idc.read_dbg_byte(addr + i)) return bytes(dd) ``` ### Read std::string (64-bit) ```python def dbg_read_cppstr_64(objectAddr): strPtr = idc.read_dbg_qword(objectAddr) result = '' i = 0 while True: onebyte = idc.read_dbg_byte(strPtr + i) if onebyte == 0: break else: result += chr(onebyte) i += 1 return result ``` ### Read C String (64-bit) ```python def dbg_read_cstr_64(objectAddr): strPtr = objectAddr result = '' i = 0 while True: onebyte = idc.read_dbg_byte(strPtr + i) if onebyte == 0: break else: result += chr(onebyte) i += 1 return result ``` ### Parse GNU C++ std::map ```python import idautils import idaapi import idc def parse_gnu_map_header(address): root = idc.read_dbg_qword(address + 0x10) return root def parse_gnu_map_node(address): left = idc.read_dbg_qword(address + 0x10) right = idc.read_dbg_qword(address + 0x18) data = address + 0x20 return left, right, data def parse_gnu_map_travel(address): # address <- std::map struct address result = [] worklist = [parse_gnu_map_header(address)] while len(worklist) > 0: addr = worklist.pop() (left, right, data) = parse_gnu_map_node(addr) if left > 0: worklist.append(left) if right > 0: worklist.append(right); result.append(data) return result # example elements = parse_gnu_map_travel(0x0000557518073EB0) for elem in elements: print(hex(elem)) ``` ### Read XMM Register (Debug) ```python def read_xmm_reg(name): rv = idaapi.regval_t() idaapi.get_reg_val(name, rv) return (struct.unpack('Q', rv.bytes())[0]) ``` ### Step Over and Wait for Debug Event ```python while ida_dbg.step_over(): wait_for_next_event(WFNE_ANY, -1) rip = idc.get_reg_value("rip") # ..... ``` ### Iterate Instructions in a Function ```python for ins in idautils.FuncItems(0x401000): print(hex(ins)) ``` ### Get Function Callees (Instruction-Based) ```python def ida_get_callees(func_addr: int) -> list: callees = [] for head in idautils.Heads(func_addr, idaapi.get_func(func_addr).end_ea): if idaapi.is_call_insn(head): callee_ea = idc.get_operand_value(head, 0) callees.append(callee_ea) return callees ``` ### Double / Complex Number Memory Operations ```python def float_to_double_bytearray(value): double_value = ctypes.c_double(value) byte_array = bytearray(ctypes.string_at(ctypes.byref(double_value), ctypes.sizeof(double_value))) return byte_array def set_pos(x, y): # complex rbp = idc.get_reg_value("rbp") complex_base = rbp - 0x260 patch_dbg_mem(complex_base, float_to_double_bytearray(x)) patch_dbg_mem(complex_base + 8, float_to_double_bytearray(y)) set_pos(5.0, 6.0) ``` --- ## Import Table ### Enumerate Import Table ```python import ida_nalt nimps = ida_nalt.get_import_module_qty() print("Found %d import(s)..." % nimps) for i in range(nimps): name = ida_nalt.get_import_module_name(i) if not name: print("Failed to get import module name for #%d" % i) name = "" print("Walking imports for module %s" % name) def imp_cb(ea, name, ordinal): if not name: print("%08x: ordinal #%d" % (ea, ordinal)) else: print("%08x: %s (ordinal #%d)" % (ea, name, ordinal)) return True ida_nalt.enum_import_names(i, imp_cb) print("All done...") ``` ### Check if Address is an Import Function ```python def ida_is_import_function(addr: int) -> bool: is_find = False nimps = ida_nalt.get_import_module_qty() for i in range(nimps): def imp_cb(ea, name, ordinal): nonlocal is_find if ea == addr: is_find = True return False return True ida_nalt.enum_import_names(i, imp_cb) return is_find ``` ### Enumerate Import Addresses ```python def ida_enum_import_addr() -> List[int]: import_addrs = [] nimps = ida_nalt.get_import_module_qty() for i in range(nimps): def imp_cb(ea, name, ordinal): nonlocal import_addrs import_addrs.append(ea) return True ida_nalt.enum_import_names(i, imp_cb) return import_addrs ``` --- ## Type Information ### Struct Member Traversal ```python def extract_struct_members(type_name): fields = [] tif = ida_typeinf.tinfo_t() if tif.get_named_type(None, type_name): offset = 0 for iter in tif.iter_struct(): # udm fsize = iter.type.get_size() fields.append({ "offset": iter.offset // 8, # bit offset "size": fsize, "type": iter.type._print() }) offset += fsize else: print(f"Unable to get {type_name} type info.") return fields extract_struct_members("sqlite3_vfs") ``` ### Enumerate All Types ```python til = ida_typeinf.get_idati() for type_name in til.get_type_names(): print(type_name) ``` ### List All Struct Types ```python def list_struct_types(): types = [] til = ida_typeinf.get_idati() for type_name in til.get_type_names(): tif = ida_typeinf.tinfo_t() if tif.get_named_type(None, type_name): if tif.is_struct(): types.append(type_name) return types ``` --- ## Hex-Rays Decompiler API ### Decompile a Function ```python # verified: IDA 9.0 dec = ida_hexrays.decompile(func_addr) # dec is an object, str(dec) converts to text print(str(dec)) ``` ### Print Microcode at Different Maturity Levels ```python def print_microcode(func_ea): maturity = ida_hexrays.MMAT_GLBOPT3 # maturity: # MMAT_ZERO, ///< microcode does not exist # MMAT_GENERATED, ///< generated microcode # MMAT_PREOPTIMIZED, ///< preoptimized pass is complete # MMAT_LOCOPT, ///< local optimization of each basic block is complete. # ///< control flow graph is ready too. # MMAT_CALLS, ///< detected call arguments # MMAT_GLBOPT1, ///< performed the first pass of global optimization # MMAT_GLBOPT2, ///< most global optimization passes are done # MMAT_GLBOPT3, ///< completed all global optimization. microcode is fixed now. # MMAT_LVARS, ///< allocated local variables hf = ida_hexrays.hexrays_failure_t() pfn = idaapi.get_func(func_ea) rng = ida_hexrays.mba_ranges_t(pfn) mba = ida_hexrays.gen_microcode(rng, hf, None, ida_hexrays.DECOMP_WARNINGS, maturity) vp = ida_hexrays.vd_printer_t() mba._print(vp) print_microcode(0x1229) ``` ### Custom Instruction to User-Defined Call ```python class udc_exit_t(ida_hexrays.udc_filter_t): def __init__(self, code, name): ida_hexrays.udc_filter_t.__init__(self) if not self.init("int __usercall %s@(int status@);" % name): raise Exception("Couldn't initialize udc_exit_t instance") self.code = code self.installed = False def match(self, cdg): return cdg.insn.itype == ida_allins.ARM_svc and cdg.insn.Op1.value == self.code def install(self): ida_hexrays.install_microcode_filter(self, True); self.installed = True def uninstall(self): ida_hexrays.install_microcode_filter(self, False); self.installed = False def toggle_install(self): if self.installed: self.uninstall() else: self.install() udc_exit = udc_exit_t(0x900001, "svc_exit") udc_exit.toggle_install() ``` ### Hexrays_Hooks ```python class MicrocodeCallback(ida_hexrays.Hexrays_Hooks): def __init__(self, *args): super().__init__(*args) def microcode(self, mba: ida_hexrays.mba_t) -> "int": print("microcode generated.") return 0 r = MicrocodeCallback() r.hook() ``` --- ## Obfuscation Helpers ### OLLVM - Set Breakpoints on Real Blocks Set breakpoints on all real block entry addresses. Real blocks are identified by finding predecessors of the OLLVM dispatcher merge point. Note: identifying real blocks by xrefs to the merge point is a heuristic and may not be fully accurate. Use IDA breakpoint groups for batch management. ```python fn = 0x401F60 ollvm_tail = 0x405D4B # OLLVM real block merge point f_blocks = idaapi.FlowChart(idaapi.get_func(fn), flags=idaapi.FC_PREDS) for block in f_blocks: for succ in block.succs(): if succ.start_ea == ollvm_tail: print(hex(block.start_ea)) idc.add_bpt(block.start_ea) ``` ### Batch Add Breakpoints ```python def brkall(l): for addr in l: idc.add_bpt(addr) ``` --- ## Firmware Helpers ### Search x86 Function Prologues and Create Functions ```python # verified: IDA 9.0 def make_x86_func(): func_headers = find_bytes_list("55 8B") for h in func_headers: idc.del_items(h) idc.create_insn(h) ida_funcs.add_func(h) ``` --- ## Basic Block Utilities ### Get Basic Block Size ```python # verified: IDA 9.0 def get_bb_size(bbaddr): fn = bbaddr f_blocks = idaapi.FlowChart(idaapi.get_func(fn), flags=idaapi.FC_PREDS) for block in f_blocks: if block.start_ea == bbaddr: return block.end_ea - block.start_ea raise Exception("Not found") ``` ### Get Basic Block by Address ```python def ida_get_bb(ea): f_blocks = idaapi.FlowChart(idaapi.get_func(ea), flags=idaapi.FC_PREDS) for block in f_blocks: if block.start_ea <= ea and ea < block.end_ea: return block return None ``` --- ## Instruction Utilities ### Search Next Instruction by Keyword ```python # verified: IDA 9.0 def search_next_insn(addr, insnkey, max_search=0x100): cnt = 0 while cnt < max_search: addr = idc.next_head(addr) dis = GetDisasm(addr) if insnkey in dis: return addr cnt += 1 return None # example # search_next_insn(addr, 'movdqa') ``` ### Undefine a Range (U key equivalent) ```python # verified: IDA 9.0 def undefine_range(start, end): for i in range(start, end): idc.del_items(i) # example # undefine_range(func_start, func_end) ``` ### Search Disassembly Text ```python # verified: IDA 9.0 def search_text_all(text): import idaapi, idc start_ea = 0 result = [] while True: start_ea = idaapi.find_text(ustr=text, x=0, y=0, sflag=idaapi.SEARCH_DOWN, start_ea=start_ea) if start_ea == idc.BADADDR: break result.append(start_ea) start_ea = idc.next_head(start_ea) return result # example for x in search_text_all('movdqa'): print(GetDisasm(x)) ``` --- ## NOP Function ```python import idaapi import idautils import idc def nop_func(addr_func, arch='arm'): func = ida_funcs.get_func(addr_func) if not func: print("Function not found!") return start = func.start_ea end = func.end_ea print(f"Nopping function at: 0x{start:x} - 0x{end:x}") if arch == 'x86': nop_bytes = [0x90] # x86 NOP elif arch == 'arm': nop_bytes = [0x1F, 0x20, 0x03, 0xD5] # ARM NOP else: print(f"Unsupported architecture: {arch}") return ea = start while ea < end: insn = ida_ua.insn_t() length = ida_ua.decode_insn(insn, ea) if length == 0: print(f"Failed to decode instruction at: 0x{ea:x}") break nop_len = len(nop_bytes) for i in range(0, length, nop_len): for j in range(nop_len): if i + j < length: idc.patch_byte(ea + i + j, nop_bytes[j]) ea += length print("Nopping complete.") # example nop_func(0x401000, 'arm') ``` --- ## IDALib (Headless IDA, IDA 9.0+) IDALib allows running IDAPython analysis scripts without opening the IDA GUI. ### Installation ```bash cd idalib/python pip install . python py-activate-idalib.py ``` ### Basic Usage ```python import idapro # must be the first import import idautils import idc # open idb/binary file ida.open_database("samples/patch.so", True) # enumerate functions for func in idautils.Functions(): func_name = idc.get_func_name(func) print("Function Name: {}, Address: {}".format(func_name, hex(func))) # close and save idb ida.close_database(save=True) ``` ### Batch Decompile to JSON ```bash Usage: decompile.py ``` decompile.py: ```python import idapro import ida_hexrays import idautils import idc import os import sys import json def _decompile_internal(): result = [] for func in idautils.Functions(): func_name = idc.get_func_name(func) print("Function Name: {}, Address: {}".format(func_name, hex(func))) dec_obj = ida_hexrays.decompile(func) if dec_obj is None: continue dec_str = str(dec_obj) result.append({ 'name': func_name, 'address': hex(func), 'decompiled': dec_str }) return result def decomple_export(file, out_file): ida.open_database(file, True) r = _decompile_internal() ida.close_database(save=False) open(out_file, "w").write(json.dumps(r, indent=4)) if __name__ == "__main__": if len(sys.argv) != 3: print("Usage: {} ".format(sys.argv[0])) sys.exit(1) decomple_export(sys.argv[1], sys.argv[2]) ``` ### Multiprocess Batch Decompile ```python import os import time from multiprocessing import Pool args = { "NUM_WORKERS": 8, "INPUT_DIR": "/Users/ctf/idek2024/baby2/baby", "OUTPUT_DIR": "/Users/ctf/idek2024/baby2/decompiled", "NUM_MAX_RETRY": 3 } def decomple_one(file, out_file): retry = 0 while True: os.system("python3 decompile.py {} {}".format(file, out_file)) if os.path.exists(out_file): break retry += 1 if retry >= args["NUM_MAX_RETRY"]: return "Failed to decompile {}".format(file) time.sleep(1) return None if __name__ == "__main__": if not os.path.exists(args["OUTPUT_DIR"]): os.makedirs(args["OUTPUT_DIR"]) files = os.listdir(args["INPUT_DIR"]) files = [os.path.join(args["INPUT_DIR"], f) for f in files] out_files = [os.path.join(args["OUTPUT_DIR"], os.path.basename(f) + ".json" ) for f in files] with Pool(args["NUM_WORKERS"]) as p: r = p.starmap(decomple_one, zip(files, out_files)) for i in r: if i is not None: print(i) ```

相关技能

Agent 工作流

Skill Creator

创造有效技能指南。 当用户想创造出新的技能(或更新现有的技能),以专业知识,工作流程,或工具集成来扩展克洛德的能力时,应该使用这种技能.

Agent 工作流

Clawdhub

使用ClawdHub CLI搜索,安装,更新并发布从taladhub.com的代理技能. 需要获取苍蝇上的新技能时使用,将安装的技能同步到最新版本或特定版本,或者发布 npm-instainddhub CLI 的新/更新的技能文件夹.

Agent 工作流

Agent Team Orchestration

管弦乐团多代理团队,任务设定周期,交接协议,审查工作流程. 使用时间: (1)建立2+特派员队伍,具有不同专业,(2)确定任务路线和生命周期(收录框_ spec_建设_审查_完成),(3)在特派员之间制定交接协议,(4)建立审查和质量关口,(5)管理特派员之间的交流和文物共享.

Agent 工作流

Superpowers

Spec-first,TDD,子代理驱动的软件开发工作流程. 当:(1)构建任何新功能或应用——触发脑暴_计划_子代理执行回路,(2)调试出一个bug或测试失败——触发系统性的根起过程,(3)用户说"让我们构建","帮助我计划","我想添加X",或"这个被打破",(4)完成一个功…