LinkedIn AI agent packs
A folder, not a document: CLAUDE.md at the root, the context files it names, and a PROMPTS/ directory. Unzip it into a Cursor workspace, attach it to a Claude or ChatGPT Project, and the assistant already knows who you sell to, what you may claim and what LinkedIn actually permits. Setup is about twenty minutes, most of which is writing down your ICP.
Every pack looks like this
linkedin-sales-agent/
├── README.md # what to fill in first
├── CLAUDE.md # the job, the reading order, the refusals
├── LINKEDIN-LIMITS.md # maintained by us — do not edit
├── ICP.md # edit this first
├── PRODUCT.md
├── OUTREACH.md
├── BRAND-VOICE.md
├── …
└── PROMPTS/
├── icp-builder.md
├── sales-navigator-search-builder.md
├── prospect-research.md
└── …Every request gets a done / next / needs handback
CLAUDE.md sets that format at the root of every pack, so an agent never just stops after answering — it says what it finished, what you do next, and what it was missing.
- CLAUDE.md
- ICP.md
- LINKEDIN-LIMITS.md
- PROMPTS/lead-qualification.md
Qualify this export of 340 Sales Navigator leads.
Scored against ICP.md, capped against LINKEDIN-LIMITS.md — 1 seat, so 100 in the "contact now" tier this week, the rest deferred or dropped.
DONE: 340 rows scored, tiered, exclusions applied. NEXT: run PROMPTS/prospect-research.md on the top 20 before you message them. NEEDS: nothing — ICP.md and LINKEDIN-LIMITS.md covered it.
7 packs
- 11 files · 8 prompts
LinkedIn Sales Agent
A working agent folder — root instruction, seven context files and the prompts — that you fill in once and use for a year.
- 7 files · 5 prompts
LinkedIn Recruiter Agent
A sourcing assistant that stays inside a 30-credit month and does not score things a profile cannot evidence.
- 6 files · 5 prompts
LinkedIn Content Agent
Drafts in your voice, built from real material, with a self-critique attached before you read them.
- 4 files · 4 prompts
LinkedIn Job Search Agent
Every message, application and profile rewrite argues the same case about you.
- 8 files · 5 prompts
LinkedIn Agency System
A per-client structure where one client’s proof point cannot end up in another client’s post.
- 12 files · 12 prompts
LinkedIn AI Starter Pack
The whole toolkit as one folder, set up in twenty minutes, whatever your role is.
- 4 files · 3 prompts
LinkedIn MCP Setup Kit
A folder you can wire an assistant to LinkedIn from, with the exposure of each route stated before the first run rather than after it.
How the packs work
What is actually in a pack?
A folder. CLAUDE.md at the root declaring the job and the order the other files should be read in, the context files it names — ICP.md, PRODUCT.md, OUTREACH.md, LINKEDIN-LIMITS.md and so on — a PROMPTS/ directory with one markdown file per prompt, and a README that tells you which two files to fill in first. Nothing is a stub: every file ships with worked example content in angle brackets.
Do I need Claude Code or Cursor?
No. Claude Code and Cursor read a CLAUDE.md at the root of a workspace automatically, which is convenient, but the same folder works as a Claude Project, a ChatGPT Project or a Gemini Gem — attach the files, paste CLAUDE.md into the instructions, done.
Why a zip rather than a document?
Because the structure is part of the content: the root file references the others by name, and the prompts expect them to exist. A single document loses that and you end up pasting a wall of text. Every pack is also offered as one concatenated .md for the cases where a zip is awkward — a locked-down laptop, or pasting straight into a Project.
How do I keep them current?
Re-download. The files are generated from the same data layer as the pricing pages, so when LinkedIn moves a limit, the next build of the pack carries the new figure and its new verification date. If you would rather not re-download, point your agent at the open dataset instead — same numbers, dated per record.
Not sure which pack?
Take the Starter Pack. It is every context file plus the twelve most-used prompts, which covers most of what the role packs contain and lets you find out which job you are actually doing.