A toolkit for AI coding agents
Just ask.
Your agent
has the tool.
Install once and your agent picks up new abilities it did not have before — no commands to memorise, no wiring. Ask in plain language and it reaches for the right one. Works in Claude Code, Codex, Cursor, Gemini CLI, Copilot and 40+ more, because it is built on an open standard.
the skill says my credential is missing
01 What's in the bag
Generated from the repository on every push. What you see here is what ships.
cloudflare-ops
v2.1.0Connect a Cloudflare account once, then manage DNS records, subdomains, Pages sites, R2 buckets and Workers from the agent without opening the dashboard again. Setup is a two-minute guided flow that needs three checkboxes, not the 390-odd permissions Cloudflare would otherwise ask you to pick by hand. Works on macOS, Windows and Linux.
dokploy-ops
v2.1.0Connect one or more self-hosted Dokploy installations once, then run what is deployed on them from the agent. See every app, database and compose stack with its real state, deploy and watch the build to the end, read logs, restart, stop, check domains and environment variables, and find out which service is broken. Setup adopts a key the machine already has, or walks through making one. Several installations are kept apart by name, so a deploy cannot land on the wrong server. Works on macOS, Windows and Linux.
elevenlabs-studio
v1.0.0Make voice-over, sound effects and music with ElevenLabs, and spend the credits on purpose. Every paid call is priced first, never paid for twice, measured after, and written to a ledger in the project. A reserve stops a runaway loop from emptying the month. Needs an ElevenLabs API key, set up once through a guided flow without the key ever being typed into the chat. Works on macOS, Windows and Linux.
hetzner-ops
v2.1.0Connect one or more Hetzner Cloud projects once, then run the servers from the agent. List them, create, start, stop, resize, delete, manage SSH keys, firewalls, volumes, private networks, DNS zones and records, and see what the project is costing — without opening the console again. Setup adopts a token the machine already has, or walks through making one in about a minute. Several projects are kept apart by name, so a change cannot land on the wrong one by accident. Works on macOS, Windows and Linux.
namecheap-ops
v2.1.0Connect one or more Namecheap accounts once, then run the domains from the agent. List them, see what expires and what will not renew by itself, check the balance, look up nameservers, check whether a name is free, see registration and renewal prices, change nameservers, and edit DNS records for domains that use Namecheap's own DNS — without opening the dashboard. Setup adopts a key the machine already has, and handles Namecheap's address whitelist, including sending requests through an SSH server with a fixed address when the machine's own address keeps changing. Several accounts are kept apart by name. Works on macOS, Windows and Linux.
toolkit-credentials
v2.1.0Shared credential setup and storage for any skill that needs an API key or token. Walks a user of any experience level through getting a credential from the provider, storing it in one standard place that works on macOS, Windows and Linux, and verifying it — without the secret ever being typed into the conversation. Other skills call this instead of inventing their own key handling.
02 Who it's for
If you already work with an AI coding agent, this is a bag of tools it can reach into. If you build them, it is a worked example of the standard.
- You use one agent
- Install once. Your agent gains abilities you can ask for in plain language, and nothing changes about how you work.
- You switch between agents
- The same tools follow you. Move from Claude Code to Cursor to Codex and your setup comes with you, unchanged.
- You work in a team
- Commit the skills to a repository and everyone who clones it gets the same tools, whichever agent each person prefers.
- You build tooling
- Every skill here is a working reference for the spec, with a validator that enforces portability before a commit lands.
03 Why it works everywhere
A skill is a folder with a SKILL.md in it. Agents look for those
folders in a shared location, so one copy serves all of them. Here is where each one actually looks.
04 Three tiers
Split by how far each one travels, not by taste.
- Skills
- 8 · every conformant agent. Instructions plus scripts, read unchanged wherever the standard is implemented.
- MCP servers
- 0 · any MCP client. Live tool surfaces with their own credentials and typed contracts.
- Plugins
- 0 · Claude Code. Subagents, hooks and commands — components the standard has no concept of.
05 Questions
What is an agent skill?
An agent skill is a folder containing a SKILL.md file: YAML frontmatter with a
name and description, followed by instructions in Markdown, plus optional
scripts/, references/ and assets/ directories. The agent loads
only the name and description at startup, and reads the full instructions only when a task matches.
That progressive loading is why an agent can keep many skills available at almost no context cost.
The format is defined by the Agent Skills specification, an open standard stewarded by the Agentic
AI Foundation and implemented by 40+ tools.
Do I need a different version of a skill for each agent?
No. A skill written to the specification is a single folder that every conformant agent reads
unchanged. The two rules that keep it portable are to reference bundled files by paths relative to the
skill root — scripts/watch.sh, never an absolute path — and to keep vendor-specific
syntax out of the instructions, since the body is read verbatim by every agent. Every skill in this
toolkit is checked against both rules automatically before it can be committed.
Where do agents look for skills on disk?
Most read ~/.agents/skills/ at the user level and .agents/skills/ inside
a project — a vendor-neutral path that several agents give precedence over their own directories.
Of the 8 agents whose documentation we verified on 2026-08-11,
7 read it. Claude Code is the exception: it reads
~/.claude/skills/, .claude/skills/ and installed plugins only, which is why
it installs from a plugin marketplace here instead.
How is this different from an MCP server?
A skill is procedural knowledge — instructions and scripts, loaded on demand, with no process running. An MCP server is a live tool surface: it runs, holds credentials, and exposes typed tools over a protocol. Reach for MCP when the capability needs a persistent connection or a typed contract; reach for a skill when a Markdown file and a shell script would do the job. This toolkit ships both, and keeps them in separate tiers because they travel differently.
Is it free, and can I use it commercially?
Yes. Everything here is MIT licensed and free to use, modify and redistribute, including in
commercial work. There is no account, no telemetry, and no API key required by the toolkit itself —
individual skills declare their own dependencies, such as ffmpeg, in a
compatibility field you can read before installing.
Built at Itqan Lab, a design and technology studio.
إتقان — itqan, the Arabic word for mastery: doing a thing precisely, and completely.