Loadouts
repeatable LinkedIn jobs your AI agent runsA Loadout is a job on a tape: write my next lead-magnet post, triage my connection requests, warm up this account. Load the tape into the deck in your LeadShark dashboard, answer a few setup questions, press play — and the agent runs the whole job, stopping for your approval before anything goes out. Prefer your own assistant? Every tape is also a portable loadout.md you can paste into Claude, ChatGPT, or Cursor.
Live today. The deck ships inside every Pro+ and Apex dashboard — and every tape’s loadout.md runs against the same Apex MCP from any agent.
The metaphor, decoded
Tape
One repeatable job — its steps, its questions, its safety rules. Written as loadout.md.
Deck
Three slots in your dashboard drawer. A loaded tape is a job that’s ready to run.
Loop
A tape at work in a slot. Run it once on demand, or set it to repeat daily/weekly at a time you pick — scheduled runs email you their result. Every run stops for your approval before acting.
What’s a Loadout?
Chatting with an AI agent is powerful but blank-page: every session starts from zero, and the good prompts live in someone’s notes. A Loadout packages the whole job instead — the steps (a see → decide → act sequence over real Apex tools), the questions it asks you before running, the approval gate it must stop at, and the rules it can never cross. Run it once or loop it daily; it behaves the same either way.
LeadShark ships a library of proven tapes (below), and you can build your own from a one-line description of the job. Both kinds run in the same deck, with the same guardrails.
Run one in LeadShark
The whole flow lives in one place: the Loadouts tab of the Apex drawer, at the bottom of your dashboard. Load → configure → play → approve. Here it is, step by step.
Open the drawer
In your LeadShark dashboard, pop open the Apex drawer at the bottom of the screen. Switch to the Loadouts tab (Agent is the default) — you’ll see the deck (3 slots) on top and the tape shelf below.
Read a tape
Tap any cassette on the shelf to read what it does: the job, the exact steps it runs, what it asks you first, and why it stays account-safe. Reading is free — nothing happens until you load and start it.
Load it into a slot
Hover the tape and hit Load (it seats itself in the first free slot) or drag it into a slot yourself. The tape just sits there, idle — loading never runs anything.
Configure
Hit ⚙ Configure and answer the tape’s few questions — audience, offer, daily caps, anything it can’t discover on its own. A tape can sit in the deck unconfigured, but it can never run until its setup is complete.
Press ▶ Start Loop
The agent runs the tape’s steps live. You see the high-level narration right under the deck — what it’s checking, what it found, what it’s drafting — so a running cassette is never a mystery.
Approve the packet
The trust momentBefore anything is sent, scheduled, or published, the run stops at the approval gate and hands you a packet: everything it found and drafted. The gate is a conversation — approve as-is, type a note and approve with changes, or hit Revise first to see the re-drafted plan before anything executes. Go back and forth until it’s right; only then do writes happen.
Keep talking to it
A finished run isn’t a dead end. Open it from Recent runs and ask a follow-up — “why these three?”, “draft one more variant” — and the agent continues with that run as context (the report modal keeps the tape’s last runs browsable, too). Every run starts with deck memory: a briefing of what all your tapes (and your own agent over MCP) did in the last week, so nothing gets double-touched — the 🧠 Memory button above the deck shows you exactly what it remembers. And the agent is yours: it’s called Finn until you rename it — and the Agent tab’s Customize view (the ⚙ in the Agent tab) lets you pick its face and color too (worn live on the drawer handle and in Agent) — and inside 🧠 Memory (or Customize’s “What the agent should know about you” — same notes, two doors) you can teach it standing notes about your business, your voice, and who to never pitch, which every future run reads (a tape’s own rules always outrank a note). For everything in between runs there’s the Agent tab (the drawer’s default): a live line to your agent that knows what’s in your deck, answers from your real account data (“how did my last post do?”), and can do quick jobs — with any write waiting for your explicit yes in the thread. Every tape your tier can run sits right above the composer as a one-tap Run chip — equipped ones play on the deck instantly, and a tape that still needs setup gets its questions asked right in the thread, one at a time, then loads itself into the deck and runs. No form. It works the other way too: hit Agent on any run report and the thread opens with that run attached, ready to be interrogated.
Loop it — or eject
Run tapes on demand, or set a Daily / Weekly loop at a time you pick (“every day at 09:00, your time” — weekly adds a day picker; timezone-aware, DST-proof). Scheduled runs email you when they stop at the gate or finish, and a waiting packet lights an amber dot on the drawer’s pull-tab. Ejecting is safe: the tape keeps its configuration and full run history, so re-inserting it later picks up exactly where you left off.
While a tape runs, its narration and any waiting approval packet appear directly under the deck — and finished runs stack below with one-line summaries. Tap any past run for the full report.
Once the deck has run, a “This week” card under the slots totals what the agent actually shipped in the last 7 days — comments, connection requests, DMs staged, posts scheduled — counted straight from run transcripts, per tape. Long jobs are fine too: a run that needs more time hands itself to a fresh worker mid-flight and keeps going, with the handoff narrated in the transcript.
Reading the deck
Each slot has a status light — the same three states you see in the drawer legend.
Looping
The tape is running or scheduled — approvals queue to you.
Loaded
Seated in a slot — either ready to run, or waiting on ⚙ Configure.
Empty
Free slot — drag a tape in, or hover one on the shelf and hit Load.
Tapes also wear their history: a small ▶ 3 odometer on a cassette is its total run count.
The drawer’s pull-tab (right edge of the dashboard) is a status light too: amber REVIEW when a run waits on your approval, green LIVE with tiny EQ bars while a run works, a blue dot for unseen activity — so you know the deck’s state without opening it.
The tape library
Nine shipped tapes across three demand loops plus an Operate layer: Create (make demand), Capture (demand that comes to you), Connect (you go to them), and Operate (keep the engine healthy). Full steps, prompts, and the loadout.md for each live in the gallery.
Create
Make demand — posts and magnets grounded in what’s working now.Capture
Demand that comes to you — inbound requests and engaged audiences, triaged into pipeline.Doorman
Pro+Every pending connection request, triaged deliberately — accepts the meaningful ones, skips the number-builders, and drafts a thoughtful welcome DM for each accept.
Scout
ApexTheir audience is your audience: pull everyone who engaged a competitor’s post, qualify them, and reach out with the context of what they just engaged with.
Interceptor
ApexFeed-native radar: patrol the live stream your network is posting, score which moments are worth being present for, and show up with the lightest right touch — like, reply, or a full comment when it warrants it.
Connect
You go to them — prospecting, warm requests, and presence, always approval-gated.Closer
ApexYour magnet captured their email — Closer finishes the job. Finds each hand-raiser on LinkedIn and picks the channel: DM if connected, warm connect if not, or email. Nothing sends without your yes.
Connector
ApexFind prospects, rank them by fit and intent, and send connection requests with real context — inside your caps, approval before anything goes out.
Wingman
ApexSee who a prospect comments on and what they react to, then show up in those rooms with substance — warm them up before any DM ever lands.
Rainmaker
ApexYour inbox is pipeline you already paid for — revive stalled DM threads with pickups that continue the real conversation, not a cold bump. Nothing sends without your yes.
Operate
The layer beneath the loops — keep automations, limits, and safety tuned.Build your own tape
The shelf has a ＋ Blank tape tile under “Your Loadouts”. Describe the job in a sentence or two — “every Friday, summarize which of my posts got the most comments and draft next week’s post” — and AI writes the full tape: steps, setup questions, approval gate, safety rules. You name it, pick an emoji and a shell color, and it lands on your shelf with your byline.
- Custom tapes run in the same deck with the same approval gate and budgets as shipped ones.
- Rename, re-emoji, or re-color any custom tape later via ✎ Edit in its drawer.
- Deleting a custom tape asks for confirmation — it removes the tape and its run history for good.
Custom tapes that act toward other people (outreach, connects, DMs) run at Apex tier, same as the shipped ones that do.
Run it in your agent
Every tape — shipped or custom — is a plain-markdown loadout.md. The .md button in a tape’s drawer copies it, and pasting it into Claude, ChatGPT, or Cursor (connected to the Apex MCP) runs the exact same job with the exact same guardrails. The deck is the easiest way to run a Loadout; the .md means you’re never locked into it.
# Creator — Loadout
```yaml
loadout: content
tier: pro-plus # Pro+ or above
mcp: https://apex.leadshark.io/mcp
```
**The job:** ...
## Before you run — ask the user
## Steps (See → Decide → Act, real tool names)
## STOP — approval gate (no writes until the user approves)
## Never (hard rules the agent can't cross)
## Done means (verifiable finish criteria)The approval gate travels with the file: whichever agent runs it must stop and show you everything before any write — and the MCP’s own limits hold regardless of what the agent does.
Why it’s safe
An agent that acts on LinkedIn as you is only worth having if you hold the reins. Loadouts inherit every guardrail Apex has — and add the approval gate on top.
Nothing sends without you
Every run stops at the approval gate before any write — posts, DMs, connects, automations. You review the packet, add feedback if you want, and approve. No approval, no action.
Your budgets are the law
Loadout runs share the same daily and weekly limits as your automations — DMs, replies, connections. A tape can never spend more than you allow.
Archived = do-not-engage
Archive a lead (Network or Leads) and Apex hard-refuses DMs to them — Loadout runs, the Drawer Agent, and MCP all share that block. Runs also check history so nobody gets double-touched.
Everything is logged
Every run, every action, every approval lands in your dashboard — recent runs show a one-line summary, and tapping one opens the full report.
Ejecting is safe
Pulling a tape out of the deck keeps its configuration and run history. Re-insert it any time and it remembers everything.
Tiers
Pro+
Runs the everyday tapes — Creator, Doorman, and Doctor: content, inbound triage, and a system health checkup. The full deck, the customizer, and every tape’s loadout.md are all included.
Apex
Unlocks the tapes that act toward other people — Closer, Connector, Scout, Interceptor, Wingman, Rainmaker — governed outreach, capture, presence, and inbox revival. An active Apex trial counts.
On Pro+, Apex tapes stay on your shelf and are fully readable — hovering shows an Apex tape · unlock pill instead of Load. Free and Pro users see the whole deck as a locked preview.
Get started
- • A LeadShark subscription — Pro+ for Creator, Doorman & Doctor, Apex (or its trial) for the outreach tapes.
- • Your LinkedIn connected in LeadShark.
- • That’s it — open the dashboard, pop the drawer, and load Creator first. One config, one play, one approval — you’ll get it immediately.
Questions or a tape you wish existed?
Tell us the job — the library grows from what operators actually run.