AI Agent Skills logopeterindia.net
Knowledge Hub

AI Agent Skills

Everything about agent skills: the open SKILL.md format and how progressive disclosure works, the coding agents and assistants whose docs describe support for it, skill collections and registries, and skill security guidance. Grouped by role, not ranked. Links go to each source's site.

26 resources•4 groups•Updated November 2026

The Agent Skills Standard

6 entries

The open format, its specification and the guidance for writing skills that trigger reliably.

01Agent Skills specificationThe complete format specification. A skill is a directory with a SKILL.md file of YAML frontmatter (name and description required) and Markdown instructions, plus optional scripts, references and assets. 02Agent Skills overviewDescribes Agent Skills as a lightweight, open format for giving agents new capabilities, loaded through progressive disclosure: discovery, activation and execution. Says the format was developed by Anthropic and released as an open standard. 03Best practices for skill creatorsGuidance on writing well-scoped skills, with companion pages on optimizing descriptions, evaluating output and using scripts. 04agentskills/agentskills (GitHub)Specification and documentation repository for Agent Skills, licensed Apache-2.0 on GitHub. 05Anthropic: Equipping agents with Agent SkillsAnthropic's introduction: organized folders of instructions, scripts and resources that give Claude procedural knowledge for real work. 06Claude Help Center: What are Skills?Anthropic's help article explaining skills for Claude users.

Agents & Tools That Support Skills

10 entries

Products whose own documentation describes support for Agent Skills. Check each product's docs for the exact folders, features and limits it supports.

07Claude CodeCreate, manage and share skills to extend Claude in Claude Code, including custom commands and bundled skills, per its docs. 08Claude (platform docs)Skills are modular capabilities that package instructions, metadata and optional resources that Claude uses when relevant, per Anthropic. 09Claude Agent SDK skillsHow to control which skills Claude can invoke in Agent SDK sessions and author skills that sessions discover. 10OpenAI Codex and ChatGPT skillsOpenAI documentation for giving ChatGPT and Codex new capabilities and expertise with skills. 11GitHub Copilot agent skillsGives Copilot repeatable workflows with task-specific instructions, scripts and resources, per GitHub Docs. 12Visual Studio Code agent skillsHow to find, adapt and create Agent Skills in VS Code for reusable workflows with supported agents. 13Cursor skillsExtends agents with specialised capabilities using Agent Skills, which Cursor calls an open standard for packaging reusable knowledge and scripts. 14Gemini CLI skillsGemini CLI documentation for Agent Skills. 15OpenCode skillsDefines reusable behaviour through SKILL.md definitions, per OpenCode docs. 16Kiro skillsExtends Kiro with portable instruction packages using the open Agent Skills standard, per its docs.

Skill Collections, Registries & Installers

8 entries

Where skills are published, searched and installed. Skills can include scripts, so review anything you install.

17anthropics/skillsAnthropic's public repository for Agent Skills. 18openai/skillsSkills catalog for Codex. 19vercel-labs/skillsThe open agent skills tool, run with npx skills, per its repository description. 20skills.shDirectory to discover and install skills for AI agents, per its site. 21ClawHubA skill registry for agents with vector search, per its site. 22VoltAgent/awesome-agent-skillsCurated collection of 1000+ agent skills from official dev teams and the community, compatible with Claude Code, Codex, Gemini CLI, Cursor and more, per its description. 23travisvn/awesome-claude-skillsCurated list of Claude Skills, resources and tools, particularly for Claude Code. 24Claude Code plugin marketplacesHow to build a plugin marketplace from a marketplace.json file and test it locally. Plugins can bundle skills with other extensions. See the Claude Code Plugins page on this site.

Skill Security & Governance

2 entries

Skills carry instructions and sometimes executable code, so they need the same scrutiny as any third-party dependency.

25OWASP Agentic Skills Top 10OWASP project that describes itself as a security framework documenting the 10 most critical security risks in AI agent skills. 26Snyk agent-scanSecurity scanner for AI agents, MCP servers and agent skills, per its repository description.

Related Pages

8 entries

Related directories on this site.

27Agentic AIThe agentic AI hub. 28Claude Code PluginsPlugins, skills, subagents, hooks and MCP for Claude Code. 29AI Coding AgentsCoding agents that load skills. 30Top MCP ServersMCP servers, the tool-connection counterpart to skills. 31AI Agent HarnessesHarnesses that run agents and their skills. 32Meta-Harness for AI AgentsTools that compose, govern and sandbox coding agents. 33AI Agent Security SolutionsSecurity tools for AI agents. 34Vulnerability Scanning Tools for AI AgentsScanners for agents, MCP servers and more.

Frequently Asked Questions

What are AI agent skills?

Agent skills are a lightweight, open format for giving AI agents new capabilities and expertise. A skill is a folder containing a SKILL.md file, with metadata and instructions that tell an agent how to perform a specific task, and optionally scripts, reference material and templates. The format was developed by Anthropic and released as an open standard, according to the Agent Skills documentation.

How do skills work?

Agents load skills through progressive disclosure in three stages. At discovery, only each skill's name and description are loaded. At activation, when a task matches a description, the agent reads the full SKILL.md. At execution, the agent follows the instructions and may run bundled scripts or load referenced files. This keeps many skills available with a small context footprint.

What does a SKILL.md file need?

According to the specification, it needs YAML frontmatter followed by Markdown. The name field is required, up to 64 characters, lowercase letters, numbers and hyphens, and must match the directory name. The description field is required, up to 1,024 characters, and should say what the skill does and when to use it. Optional fields include license, compatibility, metadata and an experimental allowed-tools list.

How are skills different from MCP servers and plugins?

Skills package instructions and know-how that an agent loads on demand, while MCP servers expose tools and data over a protocol, and plugins can bundle several kinds of extensions, sometimes including skills. They are complementary, and many setups use all three. See Top MCP Servers and Claude Code Plugins on this site.

Which agents support skills?

Anthropic's Claude products and Claude Code, OpenAI's Codex and ChatGPT, GitHub Copilot, Visual Studio Code, Cursor, Gemini CLI, OpenCode and Kiro each document support, and the open standard's own site keeps a client showcase. Details such as folder locations and supported fields differ, so check each product's documentation.

Are skills safe to install?

Not automatically. A skill can contain instructions and executable scripts, so a malicious or careless skill can cause harm. Review skills before installing them, prefer sources you trust, scan them, and apply least privilege. The OWASP Agentic Skills Top 10 and tools such as Snyk's agent-scan address skill security, and the AI Agent Security Solutions page on this site covers broader agent security.

How do I write a good skill?

The Agent Skills guidance recommends well-scoped skills, descriptions that say both what the skill does and when to use it with specific keywords, bundled scripts only where needed, and evaluating outputs with test prompts. Poor descriptions make skills trigger unreliably.

How current is this page?

Links and facts were checked on 4 November 2026 against the Agent Skills specification and each product's documentation, and all outbound links returned 200 to automated checks. This is a fast-moving area, so confirm current details in each product's docs. Descriptions follow each source's own pages and are not rankings or endorsements.