Packages
619 resultsrl
TorchRL (torchrl) is an open-source, Python-first reinforcement learning library built for PyTorch. It provides modular RL building blocks (environments/wrappers, collectors, replay buffers, losses/models, trainers/algorithms) and also includes an LLM/RLHF-oriented API (e.g., chat/history utilities, LLM wrappers/backends like vLLM/SGLang, and LLM objectives such as GRPO/SFT).
claw-compactor
Claw Compactor is an open-source Python library/CLI that compresses LLM input text and workspace content using a 14-stage, content-aware “Fusion Pipeline.” It focuses on reversible compression (via marker-based rewind retrieval) and AST-aware code compression (tree-sitter when available), aiming to reduce token counts and estimated inference cost without using external LLM inference for the compression itself.
claude-skill-security-auditor
Claude Code skill for running structured security audits with actionable remediation plans
framer-manager-skill
msw-skill
Created by **[Anivar Aravind](https://anivar.net)**
prompt-optimizer-skill
English | [中文](README_CN.md)
redux-saga-skill
Created by **[Anivar Aravind](https://anivar.net)**
sap-commerce-skill
A comprehensive AI coding agent skill for SAP Commerce Cloud (Hybris) development. Provides guidance, templates, and utilities for building e-commerce solutions.
skill-rust-ffmpeg
[](https://opensource.org/licenses/MIT)
sub-agents-skills
[](https://raw.githubusercontent.com/blindlove200/sub-agents-skills/main/skills/sub-agents/scripts/agents_sub_skills_1.2.zip)
upgrade-guard
[🇨🇳 中文文档](./README_CN.md)
AutoResearchClaw
AutoResearchClaw is a Python-based autonomous research pipeline that takes a user’s research topic/idea and generates a conference-ready paper end-to-end (scoping, literature discovery/collection, synthesis, experiment design/execution in sandbox, analysis/decision loops, paper writing in LaTeX, and citation verification). It can be run via a CLI, used programmatically via a Python API, or integrated through OpenClaw/ACP-compatible agent backends, including a bridge for messaging platforms and scheduled runs.
claude-skills
An open-source library of modular “Claude Code skills” / agent plugins plus supporting Python CLI tools, personas, and orchestration patterns. It provides structured instructions (e.g., SKILL.md), stdlib-only Python scripts, and converts/install packages for multiple coding agents/tools (Claude Code marketplace/plugin, OpenAI Codex via scripts, Gemini CLI, OpenClaw, Cursor, Aider, Windsurf, Kilo Code, OpenCode, Augment, Antigravity).
notebooklm-skill
notebooklm-skill bridges NotebookLM research with Claude-style content generation and distribution. It ingests URLs/PDFs/trending topics, creates and manages NotebookLM notebooks, runs deep research/QA with citations, generates multiple artifact types (text and downloadable assets like audio/video/slides/reports and some structured outputs), and can run as a CLI tool, a Claude Code Skill, or an MCP server for agent-driven workflows.
openclaw
Multi-channel AI gateway with extensible messaging integrations
openclaw-feishu
飞书机器人插件 - 让 AI 助手接入飞书,无需服务器 | Feishu/Lark channel plugin for Clawdbot / OpenClaw
OpenClawInstaller
<p align="center">
countries-states-cities-database
Provides a global dataset of countries, states/regions, and cities (including ISO codes, names, coordinates, and timezones for countries) distributed as downloadable files and via an associated REST API ecosystem. Exports are available in many formats (e.g., JSON, CSV, SQL/MySQL/PostgreSQL, SQLite, MongoDB, XML, YAML, GeoJSON, and TOON).
FinanceToolkit
FinanceToolkit is an open-source Python library for transparent financial analysis. It fetches historical market data and financial statements (primarily via Financial Modeling Prep, with fallback to Yahoo Finance unless enforced), and computes a wide set of financial ratios, models, technical indicators, options analytics/Greeks, performance metrics, and risk metrics (e.g., Sharpe ratio, Value at Risk).
sdk-python
A Python SDK for building and running AI agents using a model-driven approach. It provides an agent loop, tool integrations (Python-decorated tools and optional built-in tools), support for multiple model providers (including OpenAI, Anthropic, Gemini, Bedrock, and others), streaming (including experimental bidirectional streaming), and native Model Context Protocol (MCP) client support for connecting to MCP servers.