Personal interviewer coaching notes every weekday in Slack

Every weekday morning at 8am, run me a personal interviewer coaching loop across Google Calendar, Metaview, and Slack. Timezone: my local timezone. Scope: interviewers on my team (I will provide a list of interviewer emails, or default to everyone I have calendar access to).

Step 1 — Pull yesterday's meetings. For each interviewer, use Google Calendar's List Events on their primary calendar with timeMin/timeMax set to yesterday 00:00 through 23:59 in their timezone (singleEvents=true so recurring meetings expand). Identify which events were candidate interviews using lightweight heuristics: title contains words like interview, screen, phone screen, onsite, loop, technical, behavioral, or the candidate's name pattern; and/or the event has at least one external attendee (email domain does not match the interviewer's). If in doubt, favor recall over precision — it is fine to look one up and then drop it if Metaview has nothing.

Step 2 — Fetch the transcript and AI notes from Metaview. Metaview exposes its data through an MCP server, so tool names are not fixed. Call Metaview's List Available Tools once at the start of the run to discover the current catalog and their input schemas. Then, for each candidate interview from Step 1, use Call a Tool with the appropriate discovered tool names to (a) find the matching Metaview interview by candidate name, interviewer, and start time, (b) pull its transcript, and (c) pull its AI notes / scorecard summary. Match tolerantly on time (±30 min) and interviewer email. If Metaview has no record, skip that interview.

Step 3 — Score against a lightweight rubric. For each interview, produce a short internal score (do not send this raw) on: opening and rapport, coverage of the role's key competencies, question quality (open-ended, probing, non-leading), listen-to-talk ratio, and follow-up depth. Use the transcript and AI notes as evidence. Cite one or two specific moments (a question asked, a follow-up missed, a strong summary) so the feedback is concrete.

Step 4 — Draft one coaching note per interviewer (not per interview). If an interviewer ran multiple interviews yesterday, roll them up into a single note. The note should be short and human: a one-line greeting, two specific things they did well (with a brief example each), and one specific thing to try next time, phrased as a suggestion not a scold. No em dashes. No corporate hedging. Around 120–180 words total.

Step 5 — Deliver via Slack. For each interviewer with at least one scored interview, use Slack's Send a Message to DM them the coaching note (open a DM first if needed by looking up their Slack user via workspace email match). Skip anyone with no interviews yesterday, no Metaview data, or no Slack account. At the end, DM me a one-line summary of who got a note, who was skipped, and why.

Guardrails: never post to a public channel; DMs only. Never include candidate PII beyond the candidate's first name. If any step returns an error or empty result for an interviewer, log it and continue with the rest of the team — one bad interview should not block the run.