Skip to content

Data & Privacy

Junior stores the data needed to continue work, deliver replies, and operate the product.

Data Purpose
Messages Transcript display, delivery, and search
Agent history Continue turns across tools, pauses, and retries
Tasks and watches Run later or follow resource events
Artifacts Deliver files created during work
Plugin records Support features from installed plugins

Sandbox files are temporary. Junior must persist or deliver a file before the sandbox disappears.

The model receives content selected for the active conversation, including the current thread, loaded skills, and tool results. Installed memory features may add content allowed for that actor and destination.

Junior uses normalized resource events instead of storing complete webhook payloads. Secret values and unrelated private conversations do not belong in model context.

The Slack destination controls conversation visibility. Private transcripts are available only to authorized participants. Child conversations, artifacts, and plugin records inherit the parent boundary.

Read access does not grant permission to act with another user’s connected account.

  • Redaction hides sensitive content while preserving the record.
  • Expiry removes content after its retention period.
  • Purge removes the complete conversation tree.

Logs and traces use safe identifiers, counts, and error classes. They should not contain tokens, prompts, raw message bodies, or credential material.

  • Do not paste secrets, credentials, or private customer data into Slack.
  • Link to approved systems instead of copying sensitive content.
  • Treat public channels as public.

Use Dashboard for transcript access and Security Hardening for production checks.