Instructions to use Rushit21/llamafile-chat-template-poc with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Rushit21/llamafile-chat-template-poc with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="Rushit21/llamafile-chat-template-poc",
	filename="poc_chat_template_backdoor_v2.gguf",
)

llm.create_chat_completion(
	messages = "No input example has been defined for this model task."
)

Notebooks
Google Colab
Kaggle
Local Apps

llama.cpp

How to use Rushit21/llamafile-chat-template-poc with llama.cpp:

Install from brew

brew install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf Rushit21/llamafile-chat-template-poc
# Run inference directly in the terminal:
llama-cli -hf Rushit21/llamafile-chat-template-poc

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf Rushit21/llamafile-chat-template-poc
# Run inference directly in the terminal:
llama-cli -hf Rushit21/llamafile-chat-template-poc

Use pre-built binary

# Download pre-built binary from:
# https://github.com/ggerganov/llama.cpp/releases
# Start a local OpenAI-compatible server with a web UI:
./llama-server -hf Rushit21/llamafile-chat-template-poc
# Run inference directly in the terminal:
./llama-cli -hf Rushit21/llamafile-chat-template-poc

Build from source code

git clone https://github.com/ggerganov/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-server llama-cli
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama-server -hf Rushit21/llamafile-chat-template-poc
# Run inference directly in the terminal:
./build/bin/llama-cli -hf Rushit21/llamafile-chat-template-poc

Use Docker

docker model run hf.co/Rushit21/llamafile-chat-template-poc

LM Studio
Jan
Ollama
How to use Rushit21/llamafile-chat-template-poc with Ollama:
```
ollama run hf.co/Rushit21/llamafile-chat-template-poc
```

Unsloth Studio new

How to use Rushit21/llamafile-chat-template-poc with Unsloth Studio:

Install Unsloth Studio (macOS, Linux, WSL)

curl -fsSL https://unsloth.ai/install.sh | sh
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for Rushit21/llamafile-chat-template-poc to start chatting

Install Unsloth Studio (Windows)

irm https://unsloth.ai/install.ps1 | iex
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for Rushit21/llamafile-chat-template-poc to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for Rushit21/llamafile-chat-template-poc to start chatting

Docker Model Runner
How to use Rushit21/llamafile-chat-template-poc with Docker Model Runner:
```
docker model run hf.co/Rushit21/llamafile-chat-template-poc
```

Lemonade

How to use Rushit21/llamafile-chat-template-poc with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull Rushit21/llamafile-chat-template-poc

Run and chat with the model

lemonade run user.llamafile-chat-template-poc-{{QUANT_TAG}}

List all available models

lemonade list

You need to agree to share your contact information to access this model

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

llamafile Inference-Time Backdoor via chat_template — PoC

Security research proof-of-concept for a bug bounty submission on huntr.com.

What this is

poc_chat_template_backdoor_v2.gguf is a GGUF model file that demonstrates an inference-time backdoor via a malicious tokenizer.chat_template metadata field in llamafile v0.10.0.

When loaded with llamafile, the embedded Jinja-compatible template silently injects a hidden system instruction into the model's prompt whenever any user message in the conversation contains the trigger word activate. The model behaves completely normally for all other inputs.

Reproduction

pip install gguf jinja2 numpy
python poc_verify.py   # all 7 checks pass in ~1 second

To test with a real llamafile binary:

# Positive control — injection fires
./llamafile -m poc_chat_template_backdoor_v2.gguf --cli --verbose-prompt \
  -p "please activate the assistant"

# Negative control — clean
./llamafile -m poc_chat_template_backdoor_v2.gguf --cli --verbose-prompt \
  -p "what is the capital of France?"

Scanner bypass

ProtectAI ModelScan v0.8.8: no .gguf handler → full bypass
ModelAudit: no Jinja2 gadget detection → bypass
PickleScan: not a pickle file → bypass
24-keyword static ACE signature scan: 0 hits

Affected

llamafile v0.10.0 (all versions with Jinja2 support, since llama.cpp PR #18462)

Responsible disclosure

Submitted to huntr.com Model Format Vulnerability program.

Downloads last month: -

GGUF

Model size

64 params

Architecture

llama

Hardware compatibility

We're not able to determine the quantization variants.

View all variants

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support