Post-meeting execution

Reference file

Setup & Customization Checklist

setup-customization-checklist.md

Setup & Customization Checklist

1. Trigger setup

This skill runs from a meeting-completion trigger — fired when {{MEETING_NOTES_TOOL}} finishes processing a meeting involving {{MEETING_OWNER}}. One skill instance per person. Reference trigger instructions:

A meeting involving [Owner] has completed and its notes are ready.

Load and follow the <Post-meeting execution> skill in full.

Context for this run:
- Meeting owner: [Name] ([email])
- Meeting record: [meeting ID / link from the notes tool]
- Internal domain: [yourcompany.com]
  • {{MEETING_NOTES_TOOL}} is connected and produces summary + full transcript + attendees. The transcript is load-bearing — the grounding rule, qualification scoring, and next-step distillation all read from it.
  • {{MEETING_OWNER}}'s LinkedIn account is connected (Step 12).
  • {{REVIEW_CHANNEL}} exists and the owner actually reads it — this is the single surface for every approval.

2. Placeholders

  • {{MEETING_OWNER}}, {{MEETING_NOTES_TOOL}}, {{CRM}}, {{REVIEW_CHANNEL}}, {{INTERNAL_DOMAIN}}, {{FOLLOWUP_SCENARIOS_SKILL}}, {{VOICE_GUIDE_SKILL}}, {{LEAD_SCORING_SKILL}}, {{ENTERPRISE_ACV_THRESHOLD}}, {{SALES_ACTIVE_TAG}}, {{METRICS_LOG}} — all filled.
  • Map the Step 6 stage-semantics table to your funnel stage names. The table references semantics (scoping, consideration, procurement...), not your labels.
  • Write or adapt the two sub-skills before going live: {{FOLLOWUP_SCENARIOS_SKILL}} (your outcome taxonomy + per-category email frameworks) and {{VOICE_GUIDE_SKILL}} (how the owner actually writes — without it every draft sounds like a bot).
  • If you don't track meeting metrics, delete Step 3 rather than leaving {{METRICS_LOG}} unfilled.

3. Policy decisions

  • The owner sends the email, always. The skill drafts; the human copies it into their own inbox (preserves CC and threading). If you'd rather have the agent send on approval, change Step 10's closing rule deliberately — don't drift into it.
  • Deal creation is a recommendation, never an action. Even a 4/4 qualification score only produces a Yes/No question. Confirm the override criteria (authority in the room + exceptional fit + structural growth pressure) match your motion.
  • Trial stage ownership: if a product signup event sets your trial stage, keep the Step 6 warning — meetings must never set it. If you have no such event, delete the warning and treat trial like any other stage.
  • LinkedIn connection requests (Step 12) default to queued for approval. If blank-note requests to actual attendees fit your norms, flip them to auto-send — that is how the workflow ran in production.
  • Default follow-up reminder of 5 business days when the owner doesn't answer the timing question — adjust to your cycle length.

4. Behavioral invariants (do not remove)

  • Non-blocking errors never skip downstream steps; only a missing meeting record aborts the run.
  • Deal existence is never evidence of a prior meeting — the 6-month recency filter and same-day-deal rule stay, or first-touch metrics will be silently wrong.
  • Read-after-write verification on every metrics-log write; a false success is surfaced, never swallowed.
  • The grounding rule: nothing enters the email that wasn't in the call.
  • Every contact enriched, new or existing — no title, not done.
  • The full email body appears in the Slack post — no truncation, no "see task for details."
  • The next-step line: hard blockers only, one line, under 150 chars.
  • Step 12 runs in its own action, last, regardless of prior failures.