> ## Documentation Index
> Fetch the complete documentation index at: https://docs.suprsend.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Agent Skills

> Equip your AI coding agent with SuprSend Agent Skills covering workflow schemas, template variants, CLI flags, and live docs for accurate code output.

Talk to your AI coding agent about SuprSend the way you'd talk to a teammate who's already read all our docs.

Each skill is a small folder of SuprSend knowledge - JSON schemas, CLI flags, examples, gotchas - that sits next to your code. When you ask your agent to do anything SuprSend-related, it loads the right skill automatically and works from the actual spec instead of guessing.

<CardGroup cols={3}>
  <Card title="Quickstart" icon="bolt" href="/reference/agent-skill-quickstart">
    Install all four skills and ship your first AI-generated workflow in under 5 minutes.
  </Card>

  <Card title="GitHub repo" icon="github" href="https://github.com/suprsend/skills">
    Browse the source. Every skill is open and auditable.
  </Card>

  <Card title="skills.sh listing" icon="shield-check" href="https://skills.sh/suprsend/skills">
    Trust Hub audit - zero Socket alerts.
  </Card>
</CardGroup>

***

## Why use Skills

<AccordionGroup>
  <Accordion title="Without skills, agents hallucinate SuprSend code" icon="ghost">
    Agents invent `node_type` values that don't exist, make up CLI flags, and fake field names. You spend more time fixing the AI's output than you saved.
  </Accordion>

  <Accordion title="With skills, agents work from the real spec" icon="circle-check">
    Skills load the canonical JSON schemas, CLI reference, and channel content shapes into your agent's context the moment it picks up a SuprSend task - so generated output validates the first time.
  </Accordion>

  <Accordion title="Zero context bloat" icon="feather">
    Skills follow the [agentskills.io](https://agentskills.io/specification) progressive-disclosure pattern - only metadata loads at startup, full instructions load on activation, deep references load on demand.
  </Accordion>
</AccordionGroup>

***

## Available skills

Four skills ship today. Each activates automatically when your agent detects a relevant task - you don't have to remember which one to load.

<CardGroup cols={2}>
  <Card title="suprsend-workflow-schema" icon="diagram-project" href="/reference/agent-skill-workflow-schema">
    **Use when:** generating or editing workflow JSON.

    Every [workflow node](https://docs.suprsend.com/docs/workflow-nodes.md), field, and validation rule.
  </Card>

  <Card title="suprsend-template-schema" icon="file-lines" href="/reference/agent-skill-template-schema">
    **Use when:** authoring or editing template variants.

    Variant envelope, per-channel content schemas across all 9 channels, Handlebars + JSONNET.
  </Card>

  <Card title="suprsend-cli" icon="terminal" href="/reference/agent-skill-cli">
    **Use when:** pushing, syncing, or managing SuprSend assets from the terminal.

    Every CLI command, flag, and common mistake to avoid.
  </Card>

  <Card title="suprsend-docs-support" icon="book-open" href="/reference/agent-skill-docs-support">
    **Use when:** asking your agent SuprSend questions or debugging.

    A map of SuprSend docs and live `.md` endpoints your agent can fetch on demand.
  </Card>
</CardGroup>

***

## What you can build

<AccordionGroup>
  <Accordion title="Generate workflow JSON from plain English" icon="diagram-project">
    Describe a workflow in a sentence - *"send an email reminder 1 hour after cart abandonment, then SMS if no order in 24 hours"* - and your agent produces valid workflow JSON ready to push.
  </Accordion>

  <Accordion title="Author template variants across 9 channels" icon="file-lines">
    *"Build an SMS variant for the welcome template in French"*, *"create a tenant-specific welcome email"* - the template skill knows the variant envelope, per-channel content shapes, and the right brand variable syntax.
  </Accordion>

  <Accordion title="Push, sync, and manage SuprSend from your terminal" icon="terminal">
    Ask your agent things like *"push this workflow to staging and commit it"* or *"sync workflows from staging to production"* - it runs the right CLI command with the right flags the first time.
  </Accordion>

  <Accordion title="Migrate templates and seed test data" icon="wand-magic-sparkles">
    Point your agent at templates from another tool or codebase and have it convert and push them into SuprSend. Or seed users, lists, and preferences for local development.
  </Accordion>

  <Accordion title="Debug with live docs in context" icon="bug">
    Your agent fetches any SuprSend docs page on demand to answer questions accurately - no stale training data, no guesswork.
  </Accordion>
</AccordionGroup>

***

## Install

<Steps>
  <Step title="Install all four skills">
    Run this in your project folder:

    ```bash theme={"system"}
    npx skills add suprsend/skills
    ```

    The installer auto-detects which agents you have on your machine and asks where to install.
  </Step>

  <Step title="Generate a service token">
    Go to [Dashboard → Account Settings → Service Tokens](https://app.suprsend.com/en/account-settings/service-tokens) and create a token for your workspace.
  </Step>

  <Step title="Authenticate the CLI">
    <Tabs>
      <Tab title="Environment variable">
        ```bash theme={"system"}
        export SUPRSEND_SERVICE_TOKEN="your_service_token_here"
        ```
      </Tab>

      <Tab title="Saved profile">
        ```bash theme={"system"}
        suprsend login --token your_service_token_here
        suprsend profile list           # see available workspaces
        suprsend profile use --name my-workspace
        ```
      </Tab>
    </Tabs>
  </Step>

  <Step title="Verify the connection">
    ```bash theme={"system"}
    suprsend profile current
    ```

    Your agent will now push to this workspace when it runs CLI commands.
  </Step>
</Steps>

***

## How Skills work

<Steps>
  <Step title="At startup">
    Only the skill name and one-line description load - your agent knows what's available, nothing more.
  </Step>

  <Step title="When you describe a task">
    The relevant skill's full instructions load into context automatically.
  </Step>

  <Step title="On demand">
    Deep references (full schemas, node tables) load only if your agent needs them for that specific task.
  </Step>
</Steps>

<Tip>
  After installation, skills are passive - nothing extra to run. Just open your agent and start describing what you want to build.
</Tip>

***

## Supported agents

Skills work with any agent that follows the [agentskills.io](https://agentskills.io/specification) spec. The installer auto-detects which agents are on your machine and asks where to install.

<AccordionGroup>
  <Accordion title="Skill directories by agent" icon="folder-tree">
    | Agent                                                        | Skill directory     |
    | ------------------------------------------------------------ | ------------------- |
    | Cursor, Amp, Cline, Codex, Copilot, Warp, Kimi CLI, OpenCode | `.agents/skills/`   |
    | Claude Code                                                  | `.claude/skills/`   |
    | Continue                                                     | `.continue/skills/` |
    | Augment                                                      | `.augment/skills/`  |
  </Accordion>
</AccordionGroup>

<Tip>
  Commit the skills folder to your repo. Your whole team gets the same agent behavior with no extra setup.
</Tip>

***

## Security

<CardGroup cols={2}>
  <Card title="Audited" icon="shield-check" href="https://skills.sh/suprsend/skills">
    Every SuprSend skill is published through [skills.sh](https://skills.sh/suprsend/skills) and passes the Gen Agent Trust Hub audit with zero Socket alerts.
  </Card>

  <Card title="Open source" icon="github" href="https://github.com/suprsend/skills">
    Browse every skill, reference file, and version history on GitHub.
  </Card>
</CardGroup>

<Warning>
  Skills run with full agent permissions. Review the contents of any skill - ours or third-party - before using it on production code.
</Warning>