AI context file: make an agent sound like your brand

If you built a page in the last lesson, you probably noticed the copy was competent and a little bland. The kind of thing that could belong to any company in your category. That is not a weakness in the agent. It knows everything about marketing in general and nothing about you in particular. This lesson closes that gap, and it is the single most useful thing in the whole course.

The fix is one file.

The file the agent always reads

Every one of these tools has a file it loads at the start of each session, before you have typed a word. In Claude Code it is called CLAUDE.md. Codex reads AGENTS.md. Gemini’s editor uses GEMINI.md, and Cursor keeps a rules file. Same idea everywhere: a plain text file, sitting in your folder, that the agent treats as standing instructions.

You put in it the things you are tired of re-explaining. Who you are, who you sell to, how you sound, and the calls you have already made and do not want reopened. Then every draft, every page, every reply starts from your context instead of from the generic middle of the internet. Nothing here depends on you remembering to paste the right prompt. The file applies itself.

What the folder ends up looking like

The pattern that holds up is a short hub file plus a folder of depth. The hub loads every session. The depth gets read only when a task calls for it.

That tree is the whole architecture. The hub holds the always-true lines and points into docs/ when a task needs the long version, which is what keeps the always-loaded part lean. The skills folder comes later in the course, and we get to why in a moment.

What goes in the hub file

Keep it lean. A bloated file is as bad as an empty one, because the agent has to weigh everything you put in. Anthropic’s own guidance for Claude Code puts a number on it: aim under 200 lines, because long always-loaded files dilute how well the agent follows any single line. Four short sections cover it.

• Positioning. What you are, who it is for, and the one or two things you do that the alternatives do not.
• Voice. How you sound, with a line or two of do and do not. “Plain and direct, no hype, no emdashes” does more than a page of adjectives.
• Customer. Who you are talking to and what they already know, so the agent stops explaining things they understand.
• Constraints. The decisions you learned the hard way. “We do not discount.” “Never call us an agency.” The lines you would correct a new hire on twice.

Where everything else goes

As a marketer you sit on far more intelligence than 200 lines. Personas, battlecards, win/loss notes, research decks. All of it deserves a home, and the routing rule is short.

• Needed on every task, sayable in a line or two: the hub.
• Needed on some tasks, or needs real depth: its own file in docs/, with one pointer line in the hub.
• Everything else: nowhere.

The docs/ files can afford to be long, because the agent reads them only when a task calls for them. One topic per file:

• The full personas and ICP detail.
• The messaging framework.
• One battlecard per competitor.
• Win/loss notes.
• The pricing rationale. The settled one-line rule stays in the hub.
• The customer-quote ledger.
• Summaries of your research reports.
• The roadmap, dated, since it moves.

And some intelligence genuinely belongs nowhere.

• Live numbers. Pasted metrics go stale within weeks, so have the agent pull them from the source at task time.
• Generic marketing wisdom. Already in the model.
• Anything the agent can read for itself in the folder.
• Raw transcripts and report dumps. Summarize into docs/, archive the original.

One test settles most edge cases: if you would not repeat it to a capable freelancer in their first week, it does not belong in the always-loaded file.

Skills plug into the same hub

Later in the course you will save repeat routines as skills. A skill is a small folder with its own instructions file: a workflow described once, plus a line saying when to use it. That line is the trigger. The agent reads it and reaches for the skill on its own whenever a task matches. The weekly report or the launch checklist runs the way you defined it, no asking by name.

The reason to mention skills now is that they only earn their keep on top of this file. A skill says how to run a job. The context file is what makes the job come out sounding like you. A hundred borrowed skills without it produce someone else’s marketing, efficiently.

I make the longer case against stacking skills in the Build Log. One honest file first. Then a handful of skills you wrote yourself, which the last lesson covers.

Let the agent interview you

You do not have to write the file cold. The agent is a decent interviewer when you tell it to be one. Open a session in your folder and paste this:

Create my context file. Use the filename your tool reads
automatically (CLAUDE.md for Claude Code, AGENTS.md for Codex,
GEMINI.md for Gemini, or the Cursor rules file). Interview me first,
one question at a time, ten questions or fewer, covering four
areas: what we sell, who buys it, how we sound, and the calls
we have already made that must not be reopened. Push for
specifics. When my answer is vague, ask for a real example.
Then write the file: under 40 lines, four sections titled
Positioning, Voice, Customer, Constraints, using my wording
wherever you can. Finish by listing what is still thin, so I
know what to add later.

Answer the way you would brief a freelancer you trust, with real answers, not aspirational ones. The interview is the point, and the skeleton is just where the answers land. When the file is written, read every line and delete anything you would not stand behind. The agent drafted it. The judgment in it has to be yours.

A living file, not a finished one

Whatever you write today is a first draft, and that is the point. The file earns its value in the weeks after, because real work keeps teaching you what belongs in it.

The agent makes the same mistake twice, and the correction becomes a line. A draft comes back sounding off, and the voice section gains the rule that would have caught it. You take on a new kind of task, and a new constraint or a new docs/ file appears. You sharpen how you think about your own company, and the positioning line sharpens with you.

Expect to be ignored sometimes too. Every so often the agent sails past a rule that is sitting right there in the file. When that happens, move the line higher, make the wording blunter, or state the rule in two places. Plain, unambiguous phrasing sticks where soft phrasing slides.

The heavier files I run on client work are the same shell as yours, just months older and better fed. Nothing about the pattern changes as it grows.

For this course, the short file is plenty. Run the interview, or just put three real lines about your positioning and voice in the file, and build the last lesson’s page again. The difference in the first draft is the whole point.