Follow Up and Close Your Captured LinkedIn Leads with an AI Agent

# Closer — Loadout

```yaml
loadout: closer
tier: apex                # Apex tier (or an active Apex trial)
mcp: https://apex.leadshark.io/mcp
```

**The job:** Your magnet captured their email — Closer finishes the job. It finds each hand-raiser on LinkedIn and picks the channel: a DM pickup if you're connected, a warm connect referencing what they took if not, or a personal email. Every send waits for your approval.

## 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. **Which lead-magnet page's leads should I work?**
2. **What's the next step you want them to take?** (book a call, reply, buy — the close)
3. **Daily follow-up cap, and anyone off-limits?** (existing customers, do-not-engage)

## Steps

1. **See** — call `list_lead_pages`: Pick the pool — see which of your lead-magnet pages and quizzes actually captured leads worth working.
2. **See** — call `get_page_leads`: Pull the hand-raisers — everyone who opted in, with the email they gave and what they came for.
3. **Decide** — call `get_lead_activity`: Read the relationship: did they open what you sent, have you talked before, anyone flagged do-not-engage — so you follow up from context, never cold.
4. **Decide** — call `search`: Find each lead on LinkedIn and learn your relationship state — connected, 2nd-degree, or email-only. The state picks the channel.
5. **Decide** — call `get_messages_with_person`: Read the existing thread so a pickup continues where it left off — never restarts the conversation.
6. **Decide** — call `manage_activity_limits`: Preflight: check remaining budgets before drafting — no send is staged into a cap that's already spent.
7. **Decide** — call `list_actions`: Preflight: check the action log — anyone this Loadout (or any other) already touched gets skipped, never double-touched.
8. **Act** — call `send_message`: Connected? A personal DM pickup referencing what they downloaded and the last thing said — approval-gated.
9. **Act** — call `send_connection_request`: Not connected? Send the warmest connect on LinkedIn — a note referencing the thing they already took from you — approval-gated, inside your caps.
10. **Act** — call `send_email`: Email-only leads get one personal note via Apex — same approval gate, logged.

## 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

Human-in-the-loop by design: Closer drafts, you send — it never messages, connects, or emails anyone on its own. Warm by definition — these people raised their hand — and Closer reads the full history before it writes (get_lead_activity, get_messages_with_person), so no one gets a cold-sounding 'just following up.' Safety is preflight, not promise: budgets are checked (manage_activity_limits) and the action log is consulted (list_actions) before anything is drafted — then every channel gates the same way: DMs, connects, and emails all wait in the approval queue and log like every other action.

Using Apex, list my lead pages and pull the leads from {lead magnet / page}. For each, check get_lead_activity — did they open what I sent, have we talked, anyone I shouldn't touch — then find them on LinkedIn with search to learn our relationship state, and read our existing thread with get_messages_with_person so any pickup continues where it left off. Before drafting, preflight: check my remaining budgets with manage_activity_limits and my action history with list_actions so no one gets double-touched. Then act by state and queue everything for my approval: connected → a DM pickup referencing what they downloaded and our last exchange; not connected → a connection request whose note references the {magnet} they grabbed; email-only → a short personal email via Apex. Skip customers and do-not-engage — nothing sends until I say so.