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

# Agent skills

> Install Syntage's Agent Skills before building an integration with an AI coding assistant.

Syntage publishes a set of [Agent Skills](https://agentskills.io) that guide AI coding agents through integration work. The skills tell the agent which documentation to consult and which safety checks to apply; these docs remain the canonical product and API contract.

<Note>
  If you are building a Syntage integration with an AI coding assistant, install these first. An agent working without them will infer endpoints, field names, extractor options, and event types — and inferred integration code is usually plausible enough to pass review while being wrong.
</Note>

## Install

### Agent Skills clients

Install the repository with a compatible Agent Skills client:

```sh theme={null}
npx skills add syntagecom/skills
```

Select the skills relevant to your integration when prompted.

### Claude Code

Add the Syntage marketplace and install the skill bundle:

```text theme={null}
/plugin marketplace add syntagecom/skills
/plugin install syntage-skills@syntage
```

## Available skills

| Skill                       | Use it for                                                                                                                       |
| --------------------------- | -------------------------------------------------------------------------------------------------------------------------------- |
| `syntage-api`               | API environment selection, authentication, entities, collections, pagination, filters, rate limits, and integration architecture |
| `syntage-entity-onboarding` | Embedded entity onboarding, onboarding events, CSP, and sandbox testing                                                          |
| `syntage-extractions`       | Starting, tracking, scheduling, and troubleshooting asynchronous extractions                                                     |
| `syntage-webhooks`          | Webhook endpoint design, signature verification, idempotent event processing, and reconciliation                                 |

Invoke a skill explicitly when asking your agent for help:

```text theme={null}
Use $syntage-api to add a sandbox Syntage client to this application.
```

## Safety defaults

Every skill applies the same baseline:

* Start development and testing in [sandbox](/guides/api/environments), where returned data is simulated.
* Do not send production requests, mutate production integration configuration, or open a live onboarding flow without your explicit confirmation.
* Keep environment-specific API keys, signing secrets, datasource credentials, and authorizations outside source code, prompts, logs, and generated examples.
* Verify endpoints, fields, filters, extractor options, and event types against the current documentation instead of guessing.

## Client metadata

`SKILL.md` contains the portable instructions and discovery metadata shared by Agent Skills clients. Optional client adapters enhance supported products without changing those instructions:

* `agents/openai.yaml` supplies OpenAI clients such as Codex with display names, short descriptions, and suggested prompts.
* `.claude-plugin/marketplace.json` makes all four skills installable as a Claude Code plugin bundle.

Clients that do not recognize an adapter continue to use `SKILL.md`.

The skills are maintained at [github.com/syntagecom/skills](https://github.com/syntagecom/skills).