Classification
Skyelight sets a type for each new thread, such as bug, idea or feedback, and writes a one-line summary of it. Thread lists and agents show the summary in place of the start of the comment.
Types
Each workspace has its own list of types. A new workspace starts with Feedback, Idea and Bug, each with its own color and glyph.
- Anyone in the workspace can add a type, including reviewers.
- Owners and admins can rename, reorder and delete types in Configure.
- Changing the list doesn’t change the type of existing threads.
Type suggestions in the composer
When you write a new thread, a chip beside Send suggests its type. The chip shows Auto until you’ve written a few words, then updates as you type. Open the chip to see a meter showing how confident the suggestion is, and to pick any type from your workspace’s list. Once you pick a type, the chip keeps it and stops suggesting.
Type suggestions work on every plan.
Summaries
About two seconds after a thread is posted, Skyelight writes its one-line summary. If the thread has no type yet, for example because an agent created it, Skyelight sets one at the same time.
When someone changes a thread’s type by hand, Skyelight uses that correction as an example for later classifications in the workspace.
Plans
Automatic types and summaries are included on every plan, Free included.
Data sent for classification
Two AI providers do this work:
- TypeSafe suggests the type from the comment as you write it, with the page title and path.
- Anthropic’s Claude writes the summary from the thread’s text.
Neither provider receives screenshots or page context. Skyelight doesn’t use your threads to train models.
Organization owners and admins can turn automatic classification off for every workspace, on any plan, in Settings Automation. While it’s off, Skyelight sends no threads to Anthropic, and new threads arrive without a summary.