---
title: Security
description: Handle Sume CLI secrets, media outputs, paid calls, and agent automation safely.
---

The Sume CLI is designed for agent use, but CLI output can still contain
sensitive or user-owned data. Treat auth, job, and media output carefully.

## Secrets

Never print or commit:

- `SUME_API_KEY`;
- local `~/.sume-com/config.json`;
- raw provider payloads.

Use environment variables or secret managers for automation:

```bash
export SUME_API_KEY="sume_live_..."
sume account get --json
```

For local development, prefer `sume login` so the CLI can create and store a
scoped key through the
[CLI login approval](https://www.sume.com/cli/login) flow. Manual keys live in
the [API Keys dashboard](https://www.sume.com/dashboard/api-keys).

## Write and paid gates

Current write commands require confirmation flags.

| Gate | Use for |
|---|---|
| `--confirm-submit` | Non-paid writes such as job cancellation. |
| `--confirm-paid` | Provider-backed Avatar 1.0 and Avatar Video 1.0 runs that can reserve or spend credits. Image/Video/Music are API-first today — apply the same confirmation discipline in your HTTP client. |

Agents should confirm user intent, submit one bounded job first when testing,
and recover existing jobs instead of blindly retrying paid commands.

## Media URLs

Sume job results can include first-party media URLs. They are public URLs but
still user data.

When reporting results:

- summarize media counts and file types;
- use local filenames instead of full remote URLs when possible;
- redact query strings and private identifiers;
- avoid dumping large raw result payloads.

## Bug reports

Include sanitized command names, error codes, request ids, and job ids when
useful. Do not include API keys, signed URLs, full private media URLs, raw
provider payloads, emails, or workspace/user ids unless engineering explicitly
asks.