iTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
An Agent Skill is a reusable folder of instructions and optional resources that helps a compatible AI agent perform a task or follow a workflow. Its central file, SKILL.md, tells the agent what to do and when; the agent can load additional references, templates, or scripts as needed. The format is open, but discovery, installation, and runtime behavior depend on the product you use.
What is an Agent Skill?
A skill packages task-specific guidance so an agent can apply a repeatable process when the situation calls for it. It is more than a prompt: a skill is a folder built around SKILL.md, with optional supporting files such as reference documents, templates, and scripts. The Agent Skills format was developed by Anthropic and released as an open standard; compatible products can use the format, but that does not mean every product supports every feature. The Agent Skills documentation describes the format and its intended portability.
![Claude AI for Beginners Bible: [5 in 1] The Ultimate Guide to Automate Your Work, Save Hours Every Week, and Use AI for Real-World Results](https://m.media-amazon.com/images/I/415+fSJacsL._SL160_.jpg)
OpenAI describes skills as “modular instructions you can use to codify processes and conventions, such as company style guides or multi-step workflows.” In practice, a skill is useful when you want an agent to follow consistent steps without re-entering the same detailed instructions for every request.
How does an agent discover and use a skill?
Many implementations use progressive disclosure: the agent initially sees a skill’s name and description, and loads its full instructions only when a task appears relevant. This keeps the detailed guidance out of context until it is needed. The usual sequence is:
#1 Best Overall
- Discovery: The system makes the skill’s name and description available to the agent.
- Activation: When a user’s task matches the description, the agent reads
SKILL.md. - Execution: The agent follows the instructions and, where supported and appropriate, consults referenced files or runs bundled code.
The description helps determine whether a skill is selected, so make it specific: state both the task it handles and the circumstances in which it should be used. A broad or vague description can make a relevant skill harder for an agent to recognize.
How do you create an Agent Skill?
Start with a folder containing SKILL.md. Put the core procedure there and add other files only when they provide useful supporting material. The exact requirements can vary by target platform, so check that platform’s current guidance before distributing the skill.
- Create the folder and file. Use a clear folder name and add
SKILL.mdat its root. - Add metadata. Put YAML frontmatter at the beginning of
SKILL.md, including a name and description. Anthropic’s format guidance specifies a lowercase, hyphenated name of no more than 64 characters and a nonempty description of no more than 1,024 characters; confirm the rules for the implementation you plan to use. - Write the procedure. Explain the task, when the skill applies, and the steps the agent should follow. Keep instructions concrete and scoped, rather than trying to cover unrelated work.
- Add supporting resources if needed. Store substantial reference material, reusable templates, and repeatable scripts in separate files or folders when the platform supports them. OpenAI’s guide suggests organizing these as
references/,assets/, andscripts/. - Test it in the target environment. Check whether the agent recognizes the description, follows the procedure, and can access any supporting files or tools the skill expects.
A minimal folder might look like this:
my-skill/
SKILL.md
references/
scripts/
assets/Not every skill needs all these folders. Keep the main workflow in SKILL.md; separate files are most helpful when they keep lengthy references or reusable materials organized.
Do these 3 things before closing this tab:
1Fix the driver behind crashes, sound loss and screen glitches2Repair Windows errors before they cause bigger problems3Scan for outdated or missing drivers - takes under a minuteHow do you install or enable a skill?
There is no single installation procedure for every agent. Follow the instructions for the product and surface where you intend to use the skill, and check its current supported paths and requirements.
OpenAI API
OpenAI’s documentation describes creating skills through directory or ZIP upload, attaching them to hosted shell tools, or supplying them inline as a ZIP bundle. It also covers versioned bundles and validation limits. Review the documentation for the exact upload and tool requirements before deployment: OpenAI API skills guide.
Claude API
Anthropic documents custom skills uploaded through the Skills API and use with the code execution tool, where skills can be selected by skill ID. Its documentation also covers prebuilt document skills. See Anthropic’s skills documentation for current setup details.
Claude Code
Anthropic documents filesystem-based custom skills in personal ~/.claude/skills/ and project-level .claude/skills/ locations. Its public repository also describes plugin-marketplace installation for its example and document skill sets. Follow the current Claude Code skills guide for the applicable setup.
The Tool Desk
Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →Claude.ai
Anthropic documents prebuilt and uploaded custom skills for Claude.ai, with plan and code-execution requirements for custom skills. Custom skills do not automatically sync among Claude.ai, the API, and Claude Code. Check Anthropic’s Claude.ai skills help page for current availability and prerequisites.
What changes between platforms?
The file format may be shared, but how a skill is installed and what it can do are platform-specific. Before relying on one, check these differences:
Rank #4
- Compatibility: Confirm that the target agent supports the skill format and the features your skill uses.
- Discovery and installation: Products may use uploads, filesystem locations, tool attachments, or skill IDs.
- Runtime and permissions: A skill that expects network access, packages, local files, or particular tools may not work in a more restricted environment.
- Sharing and versioning: Check how the product handles updates, bundles, and whether a skill is available across different surfaces.
For example, Anthropic says Claude API skills run in a sandbox without network access or runtime package installation, while Claude Code has the network access of other programs on the user’s computer. A procedure that depends on an external service or package therefore needs different handling in those environments. Confirm the current runtime rules for your exact setup in the relevant Anthropic skills documentation.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.How should you review a skill before using it?
Treat an unfamiliar skill as untrusted until you have inspected its instructions and code. A skill can influence an agent’s planning, tool use, and command execution; malicious or careless instructions can create risks, including prompt-injection-driven data exfiltration. OpenAI’s guidance recommends reviewing skills and their supporting files before making them available to an agent: OpenAI API skills guide.
Free tools Windows power users keep installed
One-click scans. No signup required.
- Read
SKILL.mdand check whether its instructions are limited to the task it claims to handle. - Inspect scripts and other executable content before allowing them to run.
- Check what files, tools, network connections, or external services the procedure expects.
- Test in the intended environment and verify that permissions match the task.
A skill’s presence does not guarantee that its assumptions about access or execution are valid. The agent’s platform and runtime permissions determine what supporting instructions and code can actually do.
Best Value
When are Agent Skills useful?
Skills are a good fit for work that benefits from a consistent, reusable procedure: for example, applying a company style guide, following a multi-step document workflow, or carrying out a recurring task with the same checks. They are less useful when the guidance is a one-off, the workflow changes constantly, or the target agent cannot access the files or tools the instructions require.
Think of the skill as the reusable procedure and the platform as the mechanism that discovers and executes it. Before sharing a skill across products, verify compatibility and runtime behavior rather than assuming an open format guarantees identical operation.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
