AI / Superagent
Screen inbound messages before your agent reads them
Superagent screens inbound email, SMS, and WhatsApp messages delivered by Bird for prompt injection and malicious links, so your application can drop dangerous payloads, route questionable replies to a human queue, and pass clean messages to your agent.
Glissez une icône sur les points. Utilisez les touches fléchées pour déplacer une icône sélectionnée. Appuyez sur Échap pour annuler le glissement.
Receive messages on your webhook and score them with Superagent
Host an HTTPS webhook endpoint to receive inbound messages from Bird and run them through Superagent before your agent processes a reply. Bird delivers the event to your server, your worker submits the message content to Superagent, and the returned verdict determines whether the message reaches your agent, goes to a human reviewer, or gets dropped.
- Developers
- Webhooks
- Créer un webhook
- URL
- https://your-app.example.com/bird/inbound
- Événements
- email.received, sms.received, whatsapp.received
- Réponse
- whsec_...
Bird shows the webhook signing secret once on creation. Store it as an environment secret before leaving the pane.
Screen message text and embedded links across email, SMS, and WhatsAppwhile routing attachments and media through your own inspection pipeline.
Trois canaux entrants. Screen incoming email, SMS, and WhatsApp replies from one webhook endpoint by subscribing to the channels your workspace receives on.
$ email.received
$ sms.received
Quatre verdicts, trois issues. Block suspicious and dangerous messages at the perimeter, route caution verdicts to a human reviewer, and let your agent act on safe messages.
$ safe
$ caution
$ dangerous
Message text and links. Superagent inspects message bodies and URLs. Handle images, audio, video, documents, and attachments through a separate scanner before passing files to an agent.
$ text
$ https links
Mettez-le en pratique.
Poursuivez avec la documentation, les guides et les exemples sur ce sujet.