Triage LinkedIn Connection Requests with an AI Agent — Accept Deliberately, Welcome Personally

# Doorman — Loadout

```yaml
loadout: inbound-triage
tier: pro-plus            # Pro+ or above
mcp: https://apex.leadshark.io/mcp
```

**The job:** Every pending connection request, triaged deliberately — no blind accepts. Ranks requesters by fit, accepts the meaningful ones (not the number-builders), and drafts a thoughtful welcome DM for each — drafts only: you send. On Apex it also reads who's been viewing your profile for extra context.

## Before you run — ask the user

Only ask what you don't already know — check the conversation, your memory,
and the account itself first. Confirm anything you inferred in one line
instead of re-interviewing.

1. **What makes a connection meaningful for you?** (roles, industries, intent — not number-builders)
2. **Any auto-decline rules?** (spam patterns, pure sellers, anything you never accept)
3. **What should the welcome message ask or offer?** (a classic: ask why they chose to connect)

## Steps

1. **See** — call `manage_invitations`: Pull every pending connection request waiting on you.
2. **See** — call `list_signals`: Optional (Apex): read your inbound heat for extra context — profile visitors, engagers — who's already paying attention. Skip it on Pro+ and the triage still runs.
3. **Decide** — call `enrich`: Enrich each requester: role, company, intent. Judge meaningful connection vs. number-builder against your criteria.
4. **Act** — call `manage_invitations`: Accept the meaningful ones; leave or decline the rest per your rules.
5. **Act**: Draft a thoughtful welcome DM for each accept — e.g. asking why they chose to connect. Drafts only on Pro+: you send them; Apex queues sends through the approval flow.

## STOP — approval gate

Present everything you found and drafted BEFORE any write. Do not send,
schedule, publish, connect, or create automations until the user approves.
Edits beat re-drafts: apply their changes, don't start over.

## Never

- Write to LinkedIn without explicit approval this run.
- Invent facts, numbers, results, or people.
- Contact anyone on the do-not-engage list, or double-touch someone
  (check `list_actions` first).
- Retry around a refusal — report what was refused and why.

## Done means

- The user approved the drafts (aim for ≤2 edit rounds).
- Approved actions are queued or scheduled — verify via `list_actions`.
- Report back: what ran, what's staged, and where to monitor it.

## Why this stays account-safe

The gentlest people-action there is: responding to people who came to you. The only writes are invitation accepts; welcome DMs stay drafts on Pro+ — sending is yours (or Apex-tier, where sends queue for approval inside your daily budgets). Do-not-engage respected, everything logged.

Using Apex, pull my pending LinkedIn connection requests (and, if I'm on Apex, my recent inbound signals — profile views, engagers — for extra context). Enrich each requester and judge them against my criteria — I want meaningful connections, not number-builders. Accept the fits, leave the rest, and draft a short, thoughtful welcome message for each accept asking why they chose to connect. Don't send anything — give me the drafts.