Clear your Guru verification backlog in one weekly app

Build me an app my subject matter experts open every week to actually clear their Guru card verification backlog, instead of ignoring the nudge emails. The whole point is that it is a working surface: I should be able to read a stale card, decide, and clear it in a few seconds without opening Guru. It is not a digest and it is not a report.

The main view is my personal verification queue. Build it from Guru's List Unverified Cards, and enrich it with List Collections and List Folders so I can see where each card lives, List Card Verifiers so I can see who is on the hook for it, and Get Card so the row can show the actual card body inline (expandable, not just a title). For each card compute how many days past due it is: Guru expresses verificationInterval in days, so days overdue is days since the card was last verified minus its verification interval. Sort most overdue first, and show the day count prominently, plus the collection name, the folder if there is one, and the listed verifiers. Give me filters for collection and for an overdue threshold, and a search box over card titles. Staleness must be derived only from verification state and verification dates, because Guru gives us no analytics operations here: never show card view counts, search volume, or unanswered query metrics.

From any row I can take three direct actions. Verify Card confirms the card as it stands. Create Card Comment leaves a note for the team, for example asking a question before I confirm it. Replace Card Verifiers hands the card to whoever actually owns that topic now, with a picker for the new verifier. Note that Guru card IDs come in a short slug form and a long UUID form, and the comment endpoint needs the long UUID, so carry both through the handlers. After any of these succeed, refresh that row in place so the queue visibly shrinks as I work, which is most of the satisfaction of using this thing.

Each row also gets a Draft a refresh button that kicks off a background agent embedded in the app. The agent reads the current card with Get Card and produces a proposed rewrite plus a short list of what looks outdated: dead or redirected links, old pricing or plan names, renamed teams or products, references to people who appear to have left, and instructions that contradict newer cards in the same collection. It writes that draft back into the app rather than into Guru. The draft shows up on the card row as a side by side diff against the current body, with the flagged issues listed above it. I then either discard it, edit it in the app, or approve it. Only on approve does the app call Update Card with the new content and then Verify Card on the same card. Show the agent's progress on the row while it runs, and let me start drafts on several cards at once and come back to them.

Add a second tab for knowledge ops, aimed at a manager rather than an individual expert. It takes the same backlog and groups it two ways. By verifier: each person with their count of overdue cards, their worst overdue day count, and the collections they are behind in. By collection: total cards, how many are currently unverified, and the resulting percentage of that collection that is still trusted, shown as a simple bar so it is obvious which parts of the knowledge base have quietly rotted. Each person row gets a nudge button that sends them a Slack message with Slack Bot Send a Message, listing their overdue cards and the worst offender first; use Look Up User by Email to resolve the Guru verifier's email to a Slack user and Open a Conversation to DM them. The nudge is a secondary convenience, not the product.

The hard rule across the whole app: nothing writes to Guru without an explicit click from me. Verifying a card is a trust signal the entire company relies on, so there is no auto verify, no bulk verify all, and no background job that confirms cards on my behalf. The agent may only propose. Every write action should show a clear confirmation state and a plain error message if Guru rejects it, and the app should keep working sensibly for a person whose queue is empty by showing what they own and when it next comes up for review.