LinkedIn ABM Skills — operating guide

This repository is a collection of four Agent Skills by ZenABM that run a LinkedIn account-based-marketing (ABM) program end to end: plan it, design it, audit it, report it. Each skill is self-contained (SKILL.md + its own references/, assets/, scripts/) and lives under skills/. They are packaged together as one Claude Code plugin and share the ZenABM data connector (.mcp.json).

Note on loading: Claude Code does not load a plugin-root CLAUDE.md as context — plugins contribute context through their skills. This file is documentation for humans and for anyone who opens the repo directly as a project. Each skill routes itself through the description in its own SKILL.md; the table below just makes that routing legible in one place.

The skills

The ABM journey

Strategy  →  Design  →  Launch  →  Audit  →  Report
(plan)       (execute)   (LinkedIn)  (fix)     (recap)

Skills feed into each other: strategy decides what to run, execution decides what each ad says and looks like, audit grades what's live, report recaps the month. Use them together, or reach for the one that matches the request.

Routing (do-not-confuse)

• Plan a new campaign / budget → abm-strategy-planning (not audit/report — nothing is live yet).
• Build the ads from a plan → abm-campaign-execution (needs a strategy in the conversation first).
• "Are my live ads any good / what do I fix" (trailing 30 days) → linkedin-abm-audit.
• "Report on last month" (full calendar month, MoM) → linkedin-abm-report.

Operating principles

1. Use real data when connected. All four skills read live LinkedIn spend, CPC/CPM/CTR, audience size, pipeline and account engagement through the ZenABM MCP (https://app.zenabm.com/api/mcp, bundled in .mcp.json). Prefer real numbers over benchmarks; fall back to benchmarks and say so on the document.
2. Never fabricate. Do not invent metrics, TLA authors, customer stories or a strategy. Ask, or pull from ZenABM. State assumptions and footnote defaults.
3. Deliver a downloadable artifact. Every skill produces a branded, self-contained HTML document with a Download-PDF button. Design for light mode; inline CSS/JS.
4. Follow the skill, in order. Each SKILL.md is the source of truth for its steps, formulas and templates. Work through multi-step skills sequentially — present, get feedback, advance; don't dump everything at once.
5. Treat outputs as directional planning figures, not precise forecasts — say so on the document.

ZenABM connector

The skills work with smart benchmarks and no account, but shine with the user's real data. To connect: start a free trial (no credit card) at https://app.zenabm.com/signup and approve the ZenABM connector when Claude prompts. The connector is defined once at the repo root in .mcp.json and shared by every skill.

Repository layout

.
├── README.md                     # front door: what each skill does + how to install
├── CLAUDE.md                     # this file: routing + operating principles
├── LICENSE                       # MIT
├── .mcp.json                     # ZenABM data connector (shared by all skills)
├── .claude-plugin/
│   ├── plugin.json               # one plugin, bundles the four skills
│   └── marketplace.json          # makes this repo an installable marketplace
└── skills/
    ├── abm-strategy-planning/    # SKILL.md + references/
    ├── abm-campaign-execution/   # SKILL.md + references/ + assets/ + scripts/ + evals/
    ├── linkedin-abm-audit/       # SKILL.md + references/ + assets/
    └── linkedin-abm-report/      # SKILL.md + references/ + assets/