You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
[feature] Per-app feedback channel (submit/list/claim/resolve + UI button) that agents can act on #2285
Executor's built-in feedback.submit (Executor app) sends feedback to the Executor team about Executor itself. There's no standard way to collect feedback about one of my own apps: a user clicking "this is broken" in an app's UI, or an agent noting that a tool did the wrong thing. Without that, an agent can't later read the feedback and improve the app.
Proposed shape
A first-class per-app feedback channel, for example:
A framework helper, such as feedbackRouter() from apps, that an app mounts as feedback. It provides submit({message, context?, source?}), list({status?, limit?}), claim({id, by}) and resolve({id, status, resolution, deployment?}), with storage owned by the app.
An optional <FeedbackButton/> in apps/react that submits from the app UI and shows recent items with their status.
A management-API listing of open feedback across apps (for example apps.feedback.list), so one agent or schedule can work through all apps without cross-app calls, which are currently deferred.
Optionally, the dashboard could show each app's feedback next to its Tools and Skills.
Alternatives
We built this ourselves on self-host 2.0.0-beta.10 (apps@0.0.1-beta.34) and it works: a feedback table, the four tools above (no approval needed), and a small React FeedbackButton. An hourly agent claims new items, edits the app source, deploys, and resolves the item with the deployment ID. The gaps:
Each app has to vendor the same code.
It can't be added to an app whose database already exists (the schema hash check in beta.10).
Activity
RhysSullivan commentedon Oct 13, 2026
good suggestion, i think executor is missing some primitives around: