~/hackyfeed $

A simple wrapper around hugging face for asking LLMs questions in a debug environment

DroidDetective is a Python-based machine learning framework designed to analyze Android application packages (APKs) for potential malware behavior. It utilizes a Random Forest classifier trained on data from approximately 14 malware families and 100 benign apps, examining permissions specified in the APK's AndroidManifest.xml file to provide predictions on whether an app is malicious. Notably, the tool offers the ability to retrain its model with new datasets and can output results in both the CLI and JSON format for further analysis.

The "Deep Learning for Code-centric Software Vulnerability Assessment" repository provides tools and data for assessing vulnerabilities in C/C++ code using deep learning models. Its primary use case is to facilitate automated vulnerability detection through various machine learning approaches, including both graph-based and non-graph models, with provisions for multitask learning. Notable features include a structured organization of data and model scripts, as well as detailed instructions for running each model.

DeepState is a framework designed for C and C++ developers that facilitates the integration of various symbolic execution and fuzzing engines through a unified interface. It allows users to write test harnesses using a Google Test-like API, enabling seamless execution across multiple backends while supporting automatic test generation and both unit and API sequence tests. Notable features include Docker support for easy setup and pre-built examples, enhancing vulnerability detection within the development lifecycle.

DeepCode is an advanced coding tool that leverages multi-agent systems to enhance code generation. It focuses on automating coding tasks, streamlining development processes, and improving code quality by integrating intelligent agent-based architectures. Notable features include its ability to facilitate collaborative code generation and provide intelligent suggestions to developers in real-time.

DamnVulnerableShoppingLLM is a deliberately insecure large language model (LLM) tool designed for educational purposes, enabling users to interact with LLMs in a vulnerable shopping list context. Its primary use case revolves around penetration testing and security research, allowing practitioners to explore vulnerabilities by configuring the LLM endpoint and API key to retrieve the admin flag{}.

The Damn Vulnerable Model Context Protocol (DVMCP) is an educational tool designed to illustrate security vulnerabilities within Model Context Protocol (MCP) implementations. It features ten progressively challenging scenarios that demonstrate various attack vectors such as prompt injection and remote access control, aimed at helping security researchers and developers understand and mitigate these vulnerabilities. The tool runs effectively in a Docker environment, providing a structured learning experience for those interested in the security aspects of applications using LLMs.

The Damn Vulnerable LLM Agent is an educational chatbot tool designed for security researchers to explore and experiment with prompt injection attacks in ReAct agents. Built on Langchain, it specifically addresses Thought/Action/Observation injection, allowing users to engage with a simulated vulnerable chatbot environment. Notable features include support for multiple Large Language Models via environment configuration and a built-in mechanism for testing prompt injection vectors.

CWEval is a tool designed to assess both the functionality and security of code generated by large language models (LLMs) on a common set of programming tasks. Notable features include a Docker-based environment for easy setup, integration with various LLMs, and the ability to conduct simultaneous evaluations, offering parameters for model selection, sample generation, and parallel processing. This enables users to verify the integrity of LLM-generated code comprehensively.

CVE-2025-55182 is a tool designed to exploit a critical prototype pollution vulnerability in React and Next.js Server Actions, allowing for Remote Code Execution (RCE). It features an automated scanning capability via a Nuclei template and a manual exploitation script in Python, enabling the execution of arbitrary commands on vulnerable servers while extracting output effectively from response headers. This tool aims to assist security professionals in assessing the impact of this vulnerability on affected versions of React and Next.js.

Vector35/community-plugins is a collection of deprecated plugins for Binary Ninja, primarily aimed at enhancing the functionality and analysis capabilities of older versions of the tool. Key features include utilities for automating tasks, adding annotations, and integrating with AI for reverse engineering purposes. Though no longer actively maintained, the repository serves as a reference for legacy plugins compatible with Binary Ninja versions 5.1 and older.

CodeAttack is a blackbox adversarial attack model specifically designed for pre-trained programming language models, effectively generating imperceptible adversarial code samples to exploit model vulnerabilities. It leverages code structure to demonstrate the susceptibility of various models, such as CodeT5 and CodeBERT, across translation, repair, and summarization tasks in multiple programming languages. Notable features include the use of abstract syntax tree (AST) and data flow graph (DFG) constraints, along with options to attack specific tokens, enhancing both efficiency and performance drop in victim models.

Code Unlearning is a PyTorch implementation designed to erase sensitive memorization in Code Language Models (CLMs) as highlighted in the authors' ICSE'26 paper. The tool focuses on detecting sensitive memorization risks and enabling selective unlearning of specific sensitive information while preserving the model's overall performance. Key features include a systematic approach to curate a Sensitive Memorization Dataset and the ability for CLMs to forget targeted information efficiently.

Cirq is a Python library designed for the creation, manipulation, and execution of quantum circuits on quantum computers and simulators. It supports the development of circuits with flexible gate definitions and parameterization, alongside features such as circuit transformation and noise modeling, making it well-suited for applications in noisy intermediate-scale quantum (NISQ) computing. Notable integrations with tools like NumPy, SciPy, and qsim, along with cross-platform compatibility, enhance its usability for researchers and developers in the quantum computing field.

Skycards API Data Repository is an automated tool designed for fetching and storing data from the Skycards API as JSON files. Its primary use case is to ensure continuous data synchronization from multiple API endpoints, including airport, aircraft model, and airline information, utilizing Kubernetes CronJobs for scheduled data retrieval and intelligent commit practices. Notable features include smart commits that only push changes when data has been modified, support for Discord webhook notifications, and robust error handling to isolate job failures without affecting the entire operation.

Canarytokens is a network activity tracking tool designed to alert users of unauthorized actions within their environments. It provides customizable token deployment options, supports various configurations for alerting mechanisms, and enhances security monitoring by leveraging email and webhook notifications. Key features include alert throttling, detailed configuration settings, and the ability to integrate with AWS and other external services for improved management and responsiveness.

Blasphemer is an enhanced fork of the Heretic language model that is optimized for macOS, specifically Apple Silicon, providing a 55% increase in performance via advanced LoRA fine-tuning and other optimizations. Its primary use case is to efficiently remove censorship from transformer-based models while allowing users to fine-tune their models with custom knowledge and offering features such as automatic checkpointing, real-time observability, and streamlined integration with llama.cpp for GGUF conversion. Notable enhancements include high-performance abliteration, improved error handling, and a comprehensive macOS user guide.

BinAssistMCP is a comprehensive Model Context Protocol (MCP) server designed to enhance Binary Ninja's binary analysis capabilities with AI-powered reverse engineering tools. It facilitates AI-assisted tasks through dual transport support and offers an extensive suite of 44 tools, streamlining the analysis process while managing multiple binaries concurrently and providing features like guided prompts and efficient caching for improved performance.

The Binary Ninja Ollama plugin allows users to integrate a locally hosted Ollama server with Binary Ninja to leverage AI for renaming functions and variables in binaries. Its primary use case is to provide an offline, cost-free solution for enhancing code readability by automatically generating descriptive names based on AI models. Notable features include bulk renaming options for functions and variables, support for various AI models, and local data handling without third-party interference.

BAM (Binary Analysis MCPs) is a modular toolbox designed for automated binary analysis, facilitating integration with reverse engineering platforms like IDA Pro, x64dbg, and Speakeasy through a unified interface. Its modular architecture allows for easy extension and fast prototyping, making it efficient for conducting comprehensive binary analysis. The tool is ready for immediate deployment, supporting Python 3.8 and above.

Awesome Python Audio is a curated repository of Python tools, libraries, and resources tailored for various aspects of audio and music processing, including synthesis, playback, analysis, and augmentation. It features a diverse selection of utilities for audio manipulation, feature extraction, and machine learning applications, such as audioread for audio decoding, librosa for analysis, and Pedalboard for applying effects. This comprehensive catalog serves as a valuable resource for developers and researchers seeking to enhance their audio processing capabilities in Python.

AutoDroid is a Python-based automation tool designed for facilitating interactions with Android devices, enabling functionalities like bulk APK extraction, application testing across multiple devices, and malware testing against antivirus solutions. It utilizes a customizable JSON configuration file to execute ADB commands, allowing users to define device and application targets as well as scripting command blocks for streamlined operations. Notable features include support for various command sequences, variable management, and the ability to run commands on multiple devices simultaneously, making it a robust solution for Android automation tasks.

AutoCorpus is an automated tool designed for generating fuzzing corpus files using large language models, particularly effective with natural language formats like JSON and XML. It allows users to create corpus files based on prompts or existing files and supports various configurations for output management, making it suitable for enhancing fuzz testing processes. Key features include GPU acceleration for improved performance, customizable corpus generation scenarios, and a user-friendly command-line interface.

Aura Inspector is a versatile tool for auditing and testing Salesforce Experience Cloud applications, aiming to identify misconfigurations and automate testing processes. It features record discovery for both guest and authenticated users, checks for self-registration capabilities, and reveals sensitive administrative functionality through "Home URLs." The tool can be run from a command line interface with customizable options for various audit scenarios.

AudioOwl is a Python library designed for the analysis of audio files using functions from librosa and recurrent neural networks (RNN). Its primary use case is to extract critical features from music files, providing output that includes sample rate, duration, and beat times in samples. Notable features include the ability to analyze various audio formats like WAV and MP3, get waveforms, and visualize beat samples efficiently.