The most popular skills and skill registries in the Hermes ecosystem, ranked by GitHub stars. Skills are reusable capabilities that follow the agentskills.io open standard — from cybersecurity to diagram generation to creative ideation.
This page ranks the whole category by stars. For picks by what you're trying to do — with install commands, compatibility notes, and a verdict per skill — see the Hermes Skills Hub.
This framework enables autonomous engineering through subagent-driven development and a mandatory workflow of brainstorming, planning, and test-driven development. It functions as a plugin for various coding agents and CLI tools to execute complex software tasks.
A local-first design and prototyping system that provides coding-agent CLI support for tools like Hermes Agent. It features a sandboxed iframe preview for live iterations and exports artifacts to formats including HTML, PDF, and MP4.
This repository provides 818 structured skills across 34 security domains, mapped to frameworks such as MITRE ATT&CK, ATLAS, and NIST AI RMF. The library is compatible with over 26 AI platforms to enhance agent security capabilities.
Cognee serves as a self-hosted knowledge graph engine that provides agents with persistent long-term memory across sessions. It combines vector embeddings with graph-based reasoning and offers clients for Python, Rust, and TypeScript.
This skill allows agents to generate editable .drawio files from natural language descriptions using 11 diagram type presets. It includes self-check and auto-fix capabilities and can convert Mermaid text into diagram formats.
Acting as a trust layer for agent-to-agent commerce, this skill facilitates natural-language mandates, delegated permissions, and x402 payments. It includes a master router and connector skills alongside a dedicated verification and dispute resolution layer.
This tool audits and rewrites content to eliminate 62 distinct categories of detectable AI writing patterns. It is compatible with Hermes and other agents, offering modes for rewriting, detection, and editing in place.
SkillClaw enables collective skill evolution by allowing skill merging and cross-pollination across multiple agents like Hermes and OpenClaw. It includes a bilingual dashboard for inspecting session traces and monitoring skill development.
This cross-platform library provides spec-conformant agent plugins compatible with Claude Code and agentskills.io. It features 12 metaskills designed to preserve state within a local documentation folder.
This collection contains 1,153 markdown-based skills for professional and life tasks that run natively in Hermes Agent. It includes ready-to-paste exports and integrations for platforms such as Slack, Telegram, and Obsidian.
This skill set enables Hermes Agent to perform video searches, browse channels, and extract transcripts or playlist data. It requires a TranscriptAPI key and is compatible with multiple coding environments including Cursor and Windsurf.
A meta-skill plugin that automatically detects reusable patterns in active workflows to propose new skills. It generates both SKILL.md documentation and Python scaffolds, allowing users to manage detected patterns via slash commands.
This project enables Hermes Agent to generate custom thinking instructions and analytical lenses for specific tasks. It enhances output quality by reporting findings alongside an analysis of potential blind spots.
This toolkit provides type-safe support for modular packages containing instructions, resources, and scripts. It utilizes progressive disclosure to improve prompt efficiency and supports remote registries for programmatic skills.
This skill transforms rough concepts into structured design documents and implementation handoffs through Lite and Full complexity modes. It ensures quality control via a spec review gate featuring PASS/FAIL evaluation models.
This literate programming skill allows agents to weave prose, diagrams, and source code into .lit.md files. It includes a tangler to extract runnable source files and automatically syncs code edits back to the primary documentation.
Maestro provides a durable harness for long-running agents using a local-first SQLite store and Git-based memory. It organizes complex work through five distinct lane types and a hierarchical structure of Human, Supervisor, Lead, and Peer roles.
This plugin implements cross-instance shared memory using markdown files to capture high-value decisions and outcomes. It establishes a pipeline for fine-tuning and transitioning to replacement models based on recorded work history.
These interoperable skills follow the Agent Skills specification to enable interactions with CCIP, VRF, and Data Feeds. They provide specialized capabilities for integrating Hermes Agent with Chainlink oracle networks and smart contracts.
This project serves the Brazilian community by providing Hermes Agent installation guides and configuration resources in Portuguese. It features a collection of skills formatted in SKILL.md that adhere to the agentskills.io standard.
These official skills integrate FLUX.2 image and FLUX 3 video generation models directly into agent workflows. The modules include specialized capabilities for prompt optimization, cinematic inserts, and product advertisement directing.
This tool converts repositories and developer discourse into version-pinned skills with full audit trails and provenance maps. It ensures precision by linking instructions to specific file lines at pinned commits.
This registry facilitates the discovery and installation of reusable AI skills with recursive dependency resolution. It includes a Python client and CLI, alongside commerce features powered by TON payments.
This skill manages complex, multi-step tasks by providing autonomous execution with real-time progress tracking. It ensures operational safety through built-in checkpoints and rollback procedures for recovery.
A local-first workflow language for AI that executes via a single Rust binary. It audits workflows for types, permissions, and costs before execution and generates tamper-evident receipts for every run.
This skill enables the management of Nextcloud files, notes, calendars, tasks, and contacts using WebDAV and CalDAV. It utilizes HTTPS and Nextcloud App Passwords for secure communication.
A progressive MCP client featuring a socket-based daemon architecture for persistent connections. It employs three-layer lazy loading and incremental discovery of tool schemas to minimize token usage.
An intelligent skill retrieval system using a five-layer approach, including FTS5, dense embeddings, and RRF fusion. It narrows skill lists before LLM calls to reduce token consumption.
A community hub for browsing, searching, and installing skills that implements the Agentic Resource Discovery (ARD) v0.9 specification. It integrates federated search with Hugging Face and GitHub.
An ADHD-friendly personal execution OS that uses local SQLite storage for habits and study blocks. It connects to AI agents via the Model Context Protocol (MCP).
A read-only adversarial code review tool that uses three parallel subagents to hunt for bugs. It combines deterministic pre-scans with a hardened sandbox to verify findings via proof-of-concept execution.
A plugin that rewrites and scores prompts across various surfaces, including Discord, Telegram, and CLI. It supports 97 languages and can reduce token usage by 20–80% per rewrite.
This project implements multi-story task decomposition with mandatory evidence-based verification gates. It includes a Python goal engine for state management and a systematic investigation protocol for causal debugging.
A multi-agent skill for simplifying code in TypeScript, JavaScript, Python, Go, and Rust. It utilizes specialized Hygiene, Clarity, and Correctness agents with a risk-tiered approval gate.
A C# and .NET implementation of agentskills.io that integrates with Microsoft.Extensions.AI. It supports automatic registration of skill directories and progressive disclosure via SKILL.md files.
A turnkey cybersecurity lab providing 2,077 total skills and over 131 tools for forensics, network, and exploitation. It integrates 28 frameworks, including MITRE ATT&CK, OWASP, and NIST.
This project provides production-tested tools for session management, multi-session reconciliation, and worktree recovery to maintain memory across agent conversations. It belongs in this list for its ability to orchestrate parallel subagents for architectural and adversarial code reviews.
This skill automates the creation of reproducible demo GIFs for repository READMEs using vhs or Playwright. It is included for its specialized utility in optimizing GIF files under 8 MB and embedding them directly into project documentation.
This tool generates portable agent skills from plain language while ensuring they adhere to current frontmatter specifications. It is essential for maintaining skill health through CI-friendly validation and weekly monitoring for specification drift.
This CLI and MCP server enables agents to search a database of over 13,870 MCP servers and 4,384 agent skills. It belongs here as a discovery layer that provides direct installation commands for tools across platforms like Hermes and Claude Code.
This host-agnostic framework creates a self-improvement loop by analyzing session history to propose and gate skill updates behind deterministic and LLM-judge evaluators. It is unique for operating locally using the Python standard library without requiring external daemons.
This skill automates USDC payments for HTTP 402 responses, supporting protocols like Pay Link, x402, and MPP. It is a critical addition for agents requiring standardized financial transaction capabilities compatible with the agentskills.io standard.
This project provides a functional quality seal by testing agent skills for structure, functionality, and security within isolated Docker containers. It belongs in this list for producing cryptographically verifiable certificates that prove a skill actually works.
This project enables deterministic path routing and Python runtime selection across native Windows, Git Bash, and WSL environments. It is included for its ability to resolve environment-specific syntax and verify project-compatible interpreters without new installations.
This plugin allows agents to earn USDC by answering support bounties on the Base network without requiring accounts or API keys. It is featured for its decentralized approach to settling payments and managing bounty context via x402.
This skill generates multi-scale Chinese news briefings ranging from global to street-level perspectives. It belongs here for its use of semantic memory to integrate user interests into structured reports with AI-driven action suggestions.
This collection provides local and private utilities for video-to-text extraction, document OCR, and e-commerce material generation. It is included for its specialized batch processing capabilities, such as applying brand overlays to marketing assets.
This project exposes 191 tools via an MCP server to provide structured web data across 31 different platforms. It belongs in this list for offering five production-grade workflow skills compatible with Hermes, Claude, and Cursor.