Setup & Customization Checklist

1. Trigger setup

This skill runs from a LinkedIn Engagement trigger on the profile owner's connected LinkedIn account — one trigger + one skill instance per tracked person. Reference trigger instructions (keep stop-conditions in the trigger so obvious noise never loads the skill):

## Overview
This trigger monitors all engagement on [Owner]'s LinkedIn posts — comments
and reactions. This is direct inbound interest in your content and
[Owner]'s thought leadership.

## Stop Conditions
Stop immediately and take no action if the engager:
- Works at [your company] ([internal domain], or known team members by name
  — include teammates whose LinkedIn doesn't show your company as employer)
- Works at a university, college, or academic institution
- Has an academic title (Professor, Lecturer, Researcher, PhD Student, ...)
- Has no identifiable company

Otherwise, load and follow the <LinkedIn Post Engagement Handler> skill.

## Notification Routing
- Only send notifications when a meaningful action was taken (outreach
  drafted, account updated, follow-up scheduled)
- Do NOT send notifications when engagement is filtered out for lack of
  intent
- Route all notifications from this trigger ONLY to [engagement channel]

• Profile owner's LinkedIn account connected to the workspace
• Decide comments-only vs comments+reactions at the trigger level (reactions are higher volume, lower signal)
• Dedicated Slack channel per profile owner recommended

2. Placeholders

All {{...}} in SKILL.md filled — most important:

• Persona lists — the "relevant personas" and "ignore" lists are the main quality lever. Tune to your buying committee before enabling.
• {{HIGH_ACV_TAG}} — must actually exist as a tag/attribute in your workspace, and something must be setting it (the high-value path never fires otherwise).
• {{SELF_SERVE_THRESHOLD}} and {{ALERT_TIERS}} — match your segmentation and scoring model.
• Sub-skills exist: CRM hygiene, lead scoring, alert format.

3. Policy decisions

• Connection requests default to queue-for-approval (high-value path step 5). The original deployment auto-sent them with no approval — deliberate, but only for the narrow high-ACV + decision-maker path with no note attached. Switching to auto-send is a conscious opt-in; align with your risk tolerance and LinkedIn account hygiene.
• Reactions are a separate lane — a lone like only logs to CRM + account memory; the {{ALERT_TIERS}} and repeat-engagement exceptions are the only escalators. Widening this is a conscious choice — reaction volume is much higher and signal much lower than comments.
• Repeat engagers get a multi-channel sequence built and queued for review (at 2+ interactions), never sent automatically. If your org wants auto-send, that's an explicit policy change on your side.
• Lead magnet section — keep it accurate. The skill must only ever share the URL currently written in that section; when no lead magnet is active, it must say so explicitly (prevents the agent improvising a link).

4. Behavioral invariants (do not remove)

• ICP company alone is not enough — persona gate applies to every engager.
• LinkedIn engagement is the lowest-weight signal; tier cap unless stacked.
• Never move an account backwards in the funnel.
• A lone reaction never fires the high-value path or an alert, and the active-buyer check runs before every repeat-engagement escalation.
• High-value Slack post fires only when an action was taken — filtered engagement stays silent.