TOOLKIT

Digital Marketing Pro

An open-source marketing operating system as a plugin: 163 skills, 24 agents, a 12-part engagement flow, eight HTTP connectors that fire real API calls behind a typed approval gate, and an empty .mcp.json you fill yourself.

by Indranil Banerjee indranilbanerjee/digital-marketing-proMIT licence809 starscommit d866753

It is a system, not a single skill. One command runs a 61-step engagement per brand and writes fifty-odd files; per-brand state, a brand switch and an agency dashboard are built in. The README says an engagement takes about an hour of model time.

The connector layer is honest about itself: the shipped `.mcp.json` is empty, connectors are opt-in, and the eight that execute end to end (Slack, HubSpot, Klaviyo, SendGrid, Brevo, Customer.io, Mailchimp, Ahrefs) go through `:execute-action` with a uniform typed-approval gate. Twenty-five more are declared in a manifest for MCP.

It carries compliance machinery most marketing tools skip: C2PA provenance, EU AI Act Article 50 disclosure, sixteen jurisdictions, and a pre-publish `:check` command.

Connects to: nothing by default; your own MCP servers and eight HTTP connectors once configured
Access: plan-only for most skills; execution skills are approval-gated
skills: 163, 1 of them write

Connect

/plugin marketplace add indranilbanerjee/neels-plugins
/plugin install digital-marketing-pro@neels-plugins

Type this inside Claude Code; it is a slash command, not a shell command.

Requires

Claude Code (canonical), or Codex, Cursor, Copilot CLI, Antigravity, Hermes, OpenClaw
Python 3 for the 93 optional scripts
API keys per connector you enable, in environment variables

Safety review of the skill text and its scripts
Network	optional	None until a connector is configured; then HTTP calls to that vendor, with stdlib only and no third-party dependencies.
Shell	yes	93 Python scripts (claim verifier, keyword clustering, AI-tell scan, authorship check, run auditor) run locally.
Secrets	optional	Connector keys only for the connectors you enable.
Deletes	no	No delete operation documented; state lives in an output folder you can inspect.

What it exposes

163 skills, and which ones change things

Some of the 163: /digital-marketing-pro:engagement :content-engine :campaign-plan :check :execute-action :doctor :brand-setup :performance-report

Writes

execute-action, behind the typed approval gate, to the eight HTTP connectors

Pairs with

Skills that work on what it reads

How to install these skills: three routes into Claude Code, and one line to check they loaded.

Method behind it

Our B2B marketing strategy guide is the measured version of the engagement flow’s first parts: positioning, ICP and funnel, with sources. If you run the plugin’s engagement, read the guide first so you can tell which of its 61 steps are load-bearing for your brand.

What is a B2B marketing strategy? What it must exclude

Questions

Questions people ask before installing

Does it connect to anything out of the box?

No, and the README says so: the shipped .mcp.json is empty and connectors are opt-in. That is the right default; you bring the MCP servers or keys.

How is this different from claude-ads?

claude-ads is paid media only, with a control plane for live account changes. Digital Marketing Pro spans Strategy, SEO, Content, CRM and Compliance, and its execution layer is HTTP connectors to messaging and CRM tools, not ad platforms.

Is 163 skills too many?

The plugin’s own routing layer exists because of that number: every description was rewritten so the model can pick one. Start from the `:engagement` command or the `--intent` help router, not the list.