Wayfinder is a planning skill built by Matt Pocock for Claude Code, Anthropic's command-line programming tool, and it also supports other agents like Codex. It was designed to solve the problem of engineering projects that are too large to fit into a single conversation. While Matt built it for engineering use cases, Shann found it works surprisingly well for breaking down marketing campaigns.

Marketing campaigns are exactly that kind of "too big" project. You know the end goal, for example "launch the product in March", but what your offer is, which channels to use, what the creative direction should be, which decisions depend on others — none of it is decided upfront. Wayfinder turns this blank slate into a map.

The map lives on your tracker (project management tool): GitHub Issues, Linear, or even just a Markdown file. Every unresolved question becomes a decision ticket, nested under the main issue. One ticket holds exactly one question.

In a marketing context, there are four types of tickets:

- **Grilling**: The agent keeps asking follow-up questions until you make a decision. What's your positioning? What's your core offer? Who is this for?

- **Prototype**: Assets built to show you options. Three creative directions, a rough landing page wireframe, a draft marketing email.

- **Research**: The agent goes off and does the legwork itself. Competitor campaign analysis, channel performance benchmarks.

- **Task**: The grunt work needed to unblock decisions. Setting up ad accounts, requesting access to analytics.

Two of its best features are really clever.

**Fog of War**: You can anticipate certain decisions that will definitely need to be made later, but you can't even define the question yet. For example, you know you'll need an email sequence down the line, but until you lock in your offer, you can't even start on it. It stays as a fog on the map until you clear all the prerequisite decisions.

**Frontier**: These are the tickets you can work on right now, with no dependencies blocking them. Decisions block each other, just like dependencies in engineering: your offer blocks your email sequence, your audience blocks your creative direction. So the tracker directly tells you what you can finalize today.

The process is simple: clear one ticket per conversation. The answer gets written back to the ticket, and the decision list on the map grows by one line. Once the whole map is cleared, that decision list becomes your full campaign brief, and you can click into any ticket to see the reasoning behind every decision.


One word of caution after running through this process once: keep the map limited to a single campaign. Global strategy maps go out of date, because later decisions are built on assumptions from earlier decisions, and those assumptions can be overturned at any time.

Matt puts it more succinctly: prototypemaxxing, not planmaxxing. Make decisions with low-cost prototypes, not endless planning documents. Someone in the comments added: "This is way better than never-ending documents."

Installation is straightforward. For Claude Code, just run `/plugin install mattpocock-skills`. For other environments, use `npx skills@latest add mattpocock/skills`, then run `/setup-matt-pocock-skills` once to configure your tracker. The full skill list is on GitHub at github.com/mattpocock/skills.

One commenter said "Using dev tools for marketing is kind of genius." Others have shared similar practices: instead of trying to get everything done in one go, split it into separate chunks. If you cram a whole big task into a single conversation, the context gets muddled, and you can't tell what the latest point is even talking about. Open a separate thread for one clear task, stop and restart if you get stuck to reset the context before continuing.

The tool is new, but the approach to breaking down problems isn't. Split big problems down until they're small enough to act on — engineering and marketing are no different in that regard.