---
name: interview-loop-packet
description: Build a focus-area packet per interviewer, then collect and roll up debrief scores after the loop. Use this skill when: Loop scheduled — Interview loop confirmed on the calendar.
---

# Prep interviewer packets and roll up debrief scores

Goal: prep a focus-area packet per interviewer and roll up the debrief scores after the loop.

Connected tools this skill uses: notion, googlecalendar, gmail, slack.

## When to use

- Loop scheduled — Interview loop confirmed on the calendar.
- Or when the user asks directly to prep a focus-area packet per interviewer and roll up the debrief scores after the loop.

## Method

Prep the interview loop for {{candidate}}, applying for {{role}}.

1. Pull the confirmed loop for {{candidate}} from {{calendar}} and the
   scorecard for {{role}} from the agent's drive.
2. Split the scorecard's competencies across the panelists so every
   competency has exactly one owner, no gaps, no overlap.
3. Build a one-page packet per interviewer: candidate résumé, their
   assigned focus area, and 2-3 suggested questions.
4. Confirm every scorecard competency maps to exactly one interviewer
   before sending. If one is unassigned, hold the packets and flag it
   instead of sending an incomplete loop.
5. Send each packet to its interviewer {{packet_lead_time}} before their
   slot, then after each slot chase that panelist for their scorecard if
   it is not logged within {{debrief_deadline_hours}} hours.
6. Once all scores are in, roll up the combined scorecard and post it to
   {{delivery_channel}} for the hiring manager.

Never advance or reject the candidate yourself — roll up the scores and
hand the decision to the hiring manager.

## Coverage splitting

Assign every scorecard competency to exactly one interviewer before any packet goes out.

## Method
1. Pull the scorecard for the role from the agent's drive.
2. List every competency the scorecard defines.
3. Pull the confirmed panel for this loop from {{calendar}}.
4. Assign each competency to exactly one panelist, matched to their strength or past assignment pattern.
5. Check the assignment for gaps (an unassigned competency) or overlap (two owners for one competency).
6. If a gap or overlap exists, hold the packets and flag it instead of sending an incomplete split.
7. Save the finalized split as this loop's assignment record.

## Rules
- Every competency gets exactly one owner — no gaps, no overlap, no exceptions.
- Reuse the standing competency-split-by-panel-size pattern in memory when the panel size matches a past loop (set on first run, then stored in memory).
- Never send packets while any competency is unassigned.
- Update the standing split pattern in memory when a new panel size is seen.

## Output
An assignment table: one row per competency, mapped to its one interviewer, confirmed gap-free before packets go out.

## Interviewer packet format

Assemble each interviewer's résumé, focus area, and suggested questions into one packet before their slot.

## Method
1. Pull the candidate's résumé and the interviewer's assigned focus area from the coverage split.
2. Draft 2-3 suggested questions targeted at that focus area's competency.
3. Assemble the packet as one page: résumé, focus area, and the suggested questions.
4. Confirm the packet references only this interviewer's assigned competency, not the full scorecard.
5. Send the packet to the interviewer {{packet_lead_time}} before their slot.
6. Log the send time per interviewer for the debrief-chasing step to reference.

## Rules
- Keep each packet to one page — focus area and questions only, no full candidate file.
- Default lead time is {{packet_lead_time}} before the slot (set on first run, then stored in memory).
- Never send a packet before the coverage split confirms it is gap-free.
- Keep packets inside the hiring team's channel — never post candidate material broadly.

## Output
One packet per interviewer: candidate résumé, their assigned focus area, and 2-3 suggested questions, sent ahead of their slot.

## Debrief rollup

Turn scattered panel scores into one scorecard, with any missing piece called out.

## Method
1. After each slot, check whether that panelist's scorecard is logged.
2. If not logged within {{debrief_deadline_hours}} hours, chase that panelist directly.
3. Once all scores are in, pull them into one combined scorecard, grouped by competency.
4. Flag any competency where the score is missing, contradictory, or borderline.
5. Summarize the panel's overall read in two or three lines, without recommending a decision.
6. Post the rolled-up scorecard to {{delivery_channel}} for the hiring manager.

## Rules
- Chase a panelist only after {{debrief_deadline_hours}} hours (set on first run, then stored in memory) — not before.
- Roll up scores exactly as submitted — never adjust or average away a panelist's rating.
- Flag gaps and contradictions explicitly; do not smooth them into a single number.
- Never recommend advance or reject — the hiring manager makes that call, not the rollup.

## Output
A combined scorecard: one row per competency with each panelist's score, flagged gaps, and a short overall summary.

## Recurring jobs

This skill pairs with scheduled jobs. Offer to set each one up; only create the ones the user approves:

- Loop confirmation watch (Daily poll of {{calendar}} for newly confirmed interview loops) — https://eluu.ai/cookbook/interview-loop-packet/jobs-setup/loop-confirmation-watch.md