Repolode
Submit
ReversecLabs

ReversecLabs/spikee

A Python toolkit for testing LLMs, guardrails and LLM applications against prompt injection and jailbreak attacks.

View on GitHubWebsite
Stars
270
Forks
54
Open issues
4
Last commit
4 days ago

What is it?

Spikee is a toolkit for assessing how well LLMs, guardrails and LLM-powered applications hold up against prompt injection and jailbreaking. It works in two stages: it generates a test dataset from seed folders, optionally transforming payloads with plugins, and then runs that dataset against a target, judging whether each attack succeeded.

Each part of the pipeline is a Python module that can be customized. Targets adapt Spikee to the system under test, judges decide success via keyword or regex matching or by using an LLM, and attacks can generate iterative or adaptive variants during a test. Multi-turn datasets, targets and attacks are also supported.

Spikee provides a CLI for generating, testing and analysing results, and a web UI. Built-in providers cover OpenAI-compatible endpoints by default, with extras for native SDKs such as Bedrock, Azure, Ollama and Groq.

Key features

  • 01Generates test datasets from seed folders with optional payload transformations via plugins.
  • 02Runs datasets against custom or built-in targets such as LLM providers.
  • 03Judges attack success with basic keyword/regex checks or LLM-based judges.
  • 04Supports dynamic attacks, including iterative and multi-turn strategies.
  • 05Provides result analysis commands and a web UI.

Use cases

  • Red-teaming a standalone LLM against jailbreak prompts.
  • Testing a chatbot or RAG application for prompt injection exposure.
  • Evaluating how well an LLM guardrail blocks malicious inputs.
  • Comparing attack success rates across models or dataset variants.

Quick start

  1. 1

    Install spikee from PyPI

    pip install spikee
  2. 2

    Create a workspace folder

    mkdir workspace
  3. 3

    Enter the workspace and initialise it with built-in datasets and modules

    cd workspace && spikee init

Things to know

  • The base install only covers OpenAI-compatible endpoints; other providers need extras like spikee[all].
  • Some sample targets and plugins need extra third-party libraries such as torch or googletrans.

In collections

Testing frameworks & test automation tools

Topics

#prompt-injection#llm-security#red-teaming#jailbreak#llm-evaluation#pentesting#genai#llm-jailbreaks#llm-red-teaming#pentesting-tools

FAQ

What is spikee?
A Python toolkit for testing LLMs, guardrails and LLM applications against prompt injection and jailbreak attacks. Spikee is a toolkit for assessing how well LLMs, guardrails and LLM-powered applications hold up against prompt injection and jailbreaking. It works in two stages: it generates a test dataset from seed folders, optionally transforming payloads with plugins, and then runs that dataset against a target, judging whether each attack succeeded.
Is spikee free to use? What license does it have?
spikee is open source under the Apache-2.0 license. Check the license terms before using it in commercial products.
Is spikee still maintained?
Actively maintained. The last commit to spikee was on Nov 3, 2026.
What can I use spikee for?
Red-teaming a standalone LLM against jailbreak prompts. Testing a chatbot or RAG application for prompt injection exposure. Evaluating how well an LLM guardrail blocks malicious inputs. Comparing attack success rates across models or dataset variants.
How do I install spikee?
Install spikee from PyPI: pip install spikee Create a workspace folder: mkdir workspace Enter the workspace and initialise it with built-in datasets and modules: cd workspace && spikee init
Where can I get spikee?
The source code and installation instructions are on GitHub: https://github.com/ReversecLabs/spikee

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects