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

# Agent skill

> Give your agent the full integration. Install it one time, and the agent obeys the rules of this API.

You can build with an agent that writes code: Claude Code, Cursor, Codex, or
another agent that reads instructions from your repo. Give the integration to
that agent. You do not need to describe it.

The skill contains the same rules that Jenzy engineers obey:

* the money model
* idempotency
* how to get a fee quote
* how to verify a webhook signature
* the risks that cost integrators one day of work

The skill makes "integrate Jenzy payouts" into a sequence. Your agent can
obey the sequence, and it can check its own work against the sequence.

## Install the skill

```bash theme={null}
npx skills add jenzy-treasury/skills --skill hermes-mwk
```

That command installs the skill into the current project. Add `-g` to install
it for each project on your machine. Add `--agent` to select one agent:

```bash theme={null}
npx skills add jenzy-treasury/skills --skill hermes-mwk -g --agent claude-code -y
```

Claude Code reads the skill in the next session. Ask it to integrate Jenzy
payouts, or use the name `/hermes-mwk`.

<Card title="jenzy-treasury/skills" icon="github" href="https://github.com/jenzy-treasury/skills">
  The source of the skill, and the copy that your agent installs. Read it
  before you install it. It is plain markdown.
</Card>

## What is in the skill

| File                       | Contents                                                                                           |
| -------------------------- | -------------------------------------------------------------------------------------------------- |
| `SKILL.md`                 | The integration: seven steps. Each step ends with a condition that your agent can check.           |
| `reference/endpoints.md`   | Each endpoint, the payout lifecycle, the fees, the limits, and the sandbox test accounts.          |
| `reference/errors.md`      | Each error code, and when to send the request again.                                               |
| `reference/webhooks.md`    | The wrapper, the event list, how to verify a signature, and the payload fields that you can trust. |
| `scripts/smoke-sandbox.sh` | Ping, fee quote, simulated pay-in, payout, and a poll to a terminal state, against the sandbox.    |

The skill has links to these docs. It does not keep a copy of them. So an
agent that can read a URL always has the current reference.

## Test the loop before you write code

The script in the skill does the full path against the sandbox. It shows the
real response at each step. It also tests the two errors that integrators
make most often:

* The fee quote must agree with the credit exactly.
* A second request with the same `Idempotency-Key` must give the same payout,
  and must not make a second payout.

```bash theme={null}
JENZY_API_KEY=jz_live_… .claude/skills/hermes-mwk/scripts/smoke-sandbox.sh
```

The script operates only against the sandbox host. So it can spend only test
money. To change the test, add `--rail bank` or `--amount 20000`.

## What the skill does not do

The skill has the rules of this API. It does not have your business logic. It
does not choose your data model, your retry infrastructure, or your ledger
design.

The skill holds your agent to the parts where an error costs money:

* decimal money values
* one `Idempotency-Key` for each payout
* a fee quote before you send a payout
* a reversal as the correct signal, while your balance catches up