# Modiqo agent instructions

## When to use rote

Use rote when an agent has completed useful work and you want to keep the working method. Rote records the tools, checks, corrections, dependencies, and order from the successful run. It turns that evidence into an inspectable Play that an approved agent can run again with fresh inputs.

Use rote for these jobs:

- Capture a successful agent run as a reusable Play.
- Inspect what a Play reads, writes, calls, and requires before it runs.
- Run a published Play with new inputs.
- Move a working method between Claude Code, Codex, Cursor, Kimi, Hermes, Pi, and other supported harnesses.
- Share a public Play or keep a private Play inside an organization.
- Keep versions, provenance, dependencies, and execution evidence for a repeated procedure.

Do not use rote as an agent model or as a replacement for an agent harness. Do not use a Play when the task needs open-ended exploration and no proven method exists yet. Let the agent explore first. Capture the method after the run succeeds.

## How to call rote

Install the official rote CLI:

```sh
curl -fsSL https://getrote.dev/install | bash
```

Then sign in and inspect the available commands:

```sh
rote login
rote whoami
rote play search "describe the job"
rote play inspect <owner/name>
rote play run <owner/name> --param value
```

Inspect a Play before you approve its effects. Supply credentials only through rote's credential controls. Do not put secrets in a prompt, Play file, URL, or log.

## Public resources

- Product guide: https://www.modiqo.ai/agent/product.md
- Developer resources: https://www.modiqo.ai/developers
- OpenAPI specification: https://www.modiqo.ai/openapi.json
- CLI documentation: https://www.modiqo.ai/docs/09-reference.md
- Public Play feed: https://www.modiqo.ai/feeds/public-plays.v1.json
- Full agent context: https://www.modiqo.ai/llms-full.txt