Hiring funnel dashboard for every open Hireflix role

Build me an app: a hiring funnel dashboard for Hireflix that shows where candidates drop off across all of our open roles. The built in Hireflix reporting does not go deep enough, so this app is the analysis layer on top of it. It is a reporting and management surface that recruiters open and work in, not a notification tool, so do not add alerting, reminder emails, or candidate chase sequences anywhere in it.

The main view is an overview table with one row per position, sourced from the Hireflix List Positions operation. List Positions returns both active and archived positions, so filter to active by default and give me a toggle to include archived ones for historical comparison. For each position, a handler pulls that position's interviews with List Interviews for Position and computes: invited count (every interview record on the position), completed count (candidates who actually recorded), reviewed count, shortlisted count, a completion rate shown as a percentage of invited, and the median number of days from invite to recording across the candidates who did record. Derive the reviewed and shortlisted counts from whatever review status, rating, or tagging fields the interview records carry, and degrade gracefully by hiding those columns if the account does not populate them. Sort by completion rate ascending by default so the worst roles surface first, and let me sort by any column.

Alongside the counts, break out stalled invites, meaning candidates who were invited and never recorded. Group them into aging buckets of under 7 days, 7 to 14 days, and more than 14 days since the invite went out. Show the buckets per position as their own columns or a small segmented bar, and show an account wide roll up at the top of the page with total invited, total completed, overall completion rate, and the total sitting in each bucket. These buckets are a drop off metric that tells me where the funnel leaks, not an action queue, and the app must never message a candidate.

Clicking any position row opens the candidate level breakdown for that role: one row per interview with the candidate's name and email, current status, the date they were invited, the date they recorded if they did, days elapsed, and which aging bucket they fall into. Let me filter that list by status and by bucket, and show the same headline metrics for the role at the top of the detail view.

From the dashboard I want three write actions, all Hireflix operations. Invite Candidate to Interview adds a NEW candidate by email to an under filled position. Hireflix allows a given email to be invited to a given position only once and returns a DuplicateEmailError on a repeat, so treat this strictly as adding new people: no re-invite, resend, or nudge button anywhere in the app, and if an invite comes back as a duplicate, show a clear inline message that this candidate already has an interview on this position. Archive Position closes out a dead role behind a confirmation step, after which it drops out of the active view unless the archived toggle is on. Update Position renames a mislabeled role inline from the overview or the detail view.

Add a Write the pipeline brief button that kicks off a background agent. The agent reads the same position and interview data, compares the roles against each other, and writes a short brief that calls out the roles with the worst completion rates, the roles where invites are aging without recordings, and the roles that now have enough finished interviews to make a hiring decision. It posts that brief to our recruiting channel using the Slack Send a Message operation, formatted for Slack. The brief also lands back in the app, so the dashboard shows the latest one with a timestamp and keeps a short history of previous briefs I can scroll back through. Let me pick the Slack channel and remember that choice.

Two technical details to get right. First, Hireflix timestamps are Unix epoch milliseconds, and every time to complete and aging calculation depends on parsing them correctly. Second, List Interviews for Position uses cursor pagination, so every handler that computes metrics must page through all interviews for the position rather than reading only the first page, otherwise the counts and medians will be silently wrong for the busiest roles. Show a last refreshed time and give me a manual refresh control.