---
title: Agent skills | Bannx
description: Install bannx Agent Skills into Claude Code, Codex, Cursor and other agents — connected skills that drive the bannx MCP server, and standalone design skills that work with any renderer.
source: https://bannx.com/docs/integrations/skills
---

Copy page

# Agent skills

Agent Skills are folders of instructions an AI agent loads when a task calls for them. bannx publishes a set your agent can install in one command:

- **Connected skills** teach the agent to do real work in your space through the [MCP server](https://bannx.com/docs/integrations/mcp): build templates, render them at scale, keep them on brand. They need a bannx API key.
- **Standalone skills** are design craft. They work with any renderer — HTML/CSS, Figma, an image model — and need no bannx account.

## [Install](https://bannx.com/docs/integrations/skills#install)

With the open-source `skills` installer (it detects Claude Code, Codex, Cursor and other agents and asks where to install):

```bash
npx skills add https://bannx.com
```

It reads the discovery index at [`/.well-known/agent-skills/index.json`](https://bannx.com/.well-known/agent-skills/index.json), lists the skills, and verifies each download against its SHA-256 digest.

To install one by hand, download its archive (`/.well-known/agent-skills/<name>.zip`) or its `SKILL.md` (`/.well-known/agent-skills/<name>/SKILL.md`) and unpack it into your agent's skills folder, for Claude Code `~/.claude/skills/<name>/`.

## [Connected skills](https://bannx.com/docs/integrations/skills#connected-skills)

Connect the MCP server first ([setup](https://bannx.com/docs/integrations/mcp)), with an API key scoped to the tools the agent needs ([authentication](https://bannx.com/docs/authentication)).

| Skill | Use it to |
| --- | --- |
| `bannx-quickstart` | Connect an agent, check the space and credits, and pick the right tool for a job |
| `bannx-template-authoring` | Turn a brief into a branded, data-driven template, then check it by rendering |
| `bannx-render-automation` | Render a template at scale: per-recipient images, OG images, bulk jobs, signed URLs |

Renders cost credits ([rendering](https://bannx.com/docs/rendering)); the skills render a few checks, not dozens, and tell the agent to read the balance with `whoami` first.

## [Standalone skills](https://bannx.com/docs/integrations/skills#standalone-skills)

| Skill | Use it to |
| --- | --- |
| `layout-design` | Compose ads, social posts, OG images, banners, flyers and slides |
| `design-critique` | Score a rendered design before it ships and turn each issue into a concrete fix |

## [How they stay current](https://bannx.com/docs/integrations/skills#how-they-stay-current)

Connected skills hold the workflow and the judgement calls; the exact document model comes from the MCP server's own guides (`get_guide`), so a skill never teaches an outdated node or field. Each skill carries a version in its frontmatter, and the installer re-downloads a skill when its digest changes.

## [Security](https://bannx.com/docs/integrations/skills#security)

A skill is instructions loaded into your agent's context. Install skills only from origins you trust. bannx skills contain no scripts. Keep your API key in your MCP client's config or a secret store — never in a prompt, a template or a URL.