## Example Attack { #example-attack }

Imagine you build a way to run a local AI agent.

It provides an API at

```
http://localhost:8000/v1/agents/multivac
```

There's also a frontend at

```
http://localhost:8000
```

/// tip

Note that both have the same host.

///

Then using the frontend you can make the AI agent do things on your behalf.

## Exemplo de Ataque { #example-attack }

Imagine que você desenvolve uma forma de executar um agente de IA local.

Ele fornece uma API em

```
http://localhost:8000/v1/agents/multivac
```

Há também um frontend em

```
http://localhost:8000
```

/// tip | Dica

Observe que ambos têm o mesmo host.

///

Usando o frontend, você pode fazer o agente de IA executar ações em seu nome.

## チェック { #check-it }

これで、パス `/v1/` 配下へのすべてのリクエストは Flask アプリケーションが処理します。

それ以外は **FastAPI** が処理します。

実行して [http://localhost:8000/v1/](http://localhost:8000/v1/) にアクセスすると、Flask からのレスポンスが表示されます:

```txt
Hello, World from Flask!
```

さらに [http://localhost:8000/v2](http://localhost:8000/v2) にアクセスすると、FastAPI からのレスポンスが表示されます:

```JSON
{
    "message": "Hello World"
}

    public DeleteForm() {
        // Default constructor with explicit documentation
    }

    /**
     * The search query used to find the document.
     * Optional field with maximum length of 1000 characters.
     */
    @Size(max = 1000)
    public String q;

    /**
     * The document ID of the document to delete.
     * Required field for identifying the specific document.
     */
    @Required
    public String docId;

MINIMIZED_COMMENTS_REASONS_TO_EXCLUDE = {"abuse", "off-topic", "duplicate", "spam"}


class RateLimiter:
    def __init__(self) -> None:
        self.last_query_cost: int = 1
        self.remaining_points: int = 5000
        self.reset_at: datetime = datetime.fromtimestamp(0, timezone.utc)
        self.last_request_start_time: datetime = datetime.fromtimestamp(0, timezone.utc)
        self.speed_multiplier: float = 1.0

        CacheBuilder.newBuilder().expireAfterAccess(1000, MILLISECONDS).ticker(fakeTicker).build();

    cache.put(0, 10);
    cache.put(2, 30);

    fakeTicker.advance(999, MILLISECONDS);
    assertThat(cache.getIfPresent(2)).isEqualTo(30);
    fakeTicker.advance(1, MILLISECONDS);
    assertThat(cache.getIfPresent(2)).isEqualTo(30);
    fakeTicker.advance(1000, MILLISECONDS);
    assertThat(cache.getIfPresent(0)).isNull();

    /** The name of the file configuration (required, maximum 200 characters). */
    @Required
    @Size(max = 200)
    public String name;

    /** The description of the file configuration (maximum 1000 characters). */
    @Size(max = 1000)
    public String description;

    /** The file paths to crawl (required, must be valid file URIs). */
    @Required
    @UriType(protocolType = ProtocolType.FILE)

        // Default buffer size is 1000; offer more than that
        for (int i = 0; i < 1100; i++) {
            helper.offer(new LogNotificationEvent(i, "ERROR", "org.test", "msg" + i, null));
        }
        List<LogNotificationEvent> events = helper.drainAll();
        // Should be capped at buffer size (1000)
        assertTrue(events.size() <= 1000);
        assertTrue(events.size() > 0);
    }

Imaginez que vous mettiez au point un moyen d’exécuter un agent IA local.

Il expose une API à l’adresse

```
http://localhost:8000/v1/agents/multivac
```

Il y a aussi un frontend à l’adresse

```
http://localhost:8000
```

/// tip | Astuce

Notez qu’ils ont le même hôte.

///

Vous pouvez alors, via le frontend, amener l’agent IA à effectuer des actions en votre nom.

export CI=true
export MINIO_KMS_AUTO_ENCRYPTION=on
export MINIO_KMS_SECRET_KEY=my-minio-key:OSMM+vkKUTCvQs9YL/CVMIMt43HFhkUpqJxTmGl6rYw=

(minio server http://localhost:9000/tmp/xl/{1...10}/disk{0...1} 2>&1 >/dev/null) &
pid=$!

export MC_HOST_myminio="http://minioadmin:minioadmin@localhost:9000/"

./mc ready myminio

./mc admin user add myminio/ minio123 minio123
./mc admin user add myminio/ minio12345 minio12345