If your marketing work feels busy but the qualified pipeline doesn’t move, the constraint usually isn’t creativity. It’s consistency. Teams write a strategy, file it away, then go back to making one-off decisions in Slack and sprint boards.
Emily Kramer made a blunt observation in her March 19, 2026 guide on building a marketing strategy skill in Claude: many marketers are getting outputs from Claude based on inadequate inputs, and the root cause is a missing (or unused) strategy foundation. Claude isn’t the fix by itself. The fix is giving Claude a strategy it can reliably reference—then forcing the team to use it.
Here’s the one move that matters: build a dedicated /marketing-strategy skill in Claude anchored to a single file (marketing-strategy.md), and make it the default input to every brief, review, and prioritization decision. Not a doc you “finish.” A system you operate.
That’s the promise. The risk is obvious too: if the file becomes another dumping ground, it’ll rot like every other strategy doc. The procedure below is designed to prevent that.
The tactic: turn strategy into a file Claude can’t ignore
Kramer’s framing is practical: marketers often skip foundational steps, then wonder why outputs don’t land. Claude will happily generate a campaign plan, positioning, or a lifecycle email sequence—whether or not the underlying choices are coherent. That’s not intelligence. That’s compliance.
The better approach is to make strategy a referenced artifact, not a quarterly project. In Kramer’s process, the center of gravity is a single file named marketing-strategy.md that holds the strategic decisions (ICP priorities, positioning, perceptions, revenue levers, and the “big bets”). Claude should be prompted to consult it every time.
One detail that’s easy to miss matters a lot in practice: Kramer recommends keeping CLAUDE.md concise (under ~200 lines) and putting the heavier strategy content in the separate /marketing-strategy skill file. The reason is operational, not aesthetic. A short global instruction set stays readable and stable; the strategy file can evolve without turning every prompt into archaeology.
This is also where the Zeigarnik effect works in your favor. When the strategy is a living file, it creates a productive kind of “unfinishedness”: every campaign review naturally asks, “Did this match the ICP and positioning we chose?” That loop closes only when the file is updated or the plan is corrected. Memory turns into process.
Step-by-step: build the file by running seven strategy exercises
Kramer’s guide outlines seven exercises to populate the /marketing-strategy skill. Don’t treat them like worksheets. Treat them like decisions you’re locking in—until the next time evidence forces a change.
Step 1: Create the central file. Start with marketing-strategy.md. Prompting can be as simple as Kramer’s example: “Create a Claude skill called /marketing-strategy.” The outcome is a stable place where strategy lives and can be referenced in other work.
Step 2: Populate it using the seven exercises. The sequence matters because later sections depend on earlier choices. Company context before ICP. ICP before positioning. Positioning before big campaigns.
- Company overview: business model, audience, competitive context, and what’s actually unique.
- ICP prioritization: a ranked list of segments, including maturity levels and how time should be allocated across them.
- Marketing advantages: the real strengths that can accelerate growth (not aspirational ones).
- Perceptions: 3–5 narratives that exist in the audience’s head—whether the team likes them or not.
- Positioning: the answers to the hard questions about differentiation and messaging.
- Revenue levers: what will drive revenue (new logos, expansion, retention, pricing) and current maturity.
- Big bet campaigns: 1–3 major campaigns designed to exploit the advantages and target the prioritized ICPs.
Step 3: Use it daily. Kramer’s point is unglamorous and correct: strategy documents often exist but aren’t used consistently, which produces random marketing efforts. The daily behavior change is the whole game—referencing the skill in briefs, campaign reviews, and prioritization, then updating it based on feedback and what changes in the business.
There’s another layer here for ops-minded teams: versioning. Kramer recommends using GitHub for version control and sharing skills across the team. That turns “strategy changed” from a vague feeling into a tracked diff. It also makes cross-functional handoffs cleaner because Sales Ops or RevOps can point to a commit, not a meeting memory.
Run it this week: a 90-minute implementation that creates real guardrails
Here’s the 5-minute version you can run this week:
- Audience: Marketing Ops owner (Priya’s role), one demand gen lead, one product marketing partner, one RevOps partner for measurement guardrails.
- Timeline: 90 minutes to stand up the file; 60 minutes later in the week to run the first two exercises; 20 minutes weekly to maintain.
- Tools: Claude + a GitHub repo (private) for versioning, as Kramer suggests.
Setup: Create marketing-strategy.md and add a table of contents with the seven exercise headings. Keep CLAUDE.md short and point it to the strategy file (the “under ~200 lines” constraint is straight from Kramer’s recommendation).
Launch: Run two exercises immediately: Company overview and ICP prioritization. Don’t try to do all seven in one sitting. That’s how teams end up with a “complete” doc they never revisit.
The hypothesis (make it falsifiable): If the team routes every campaign brief and review through a /marketing-strategy skill anchored to marketing-strategy.md, then execution quality will improve (fewer off-ICP campaigns and fewer positioning contradictions) because Claude will force consistent inputs and surface mismatches.
Success = directional reduction in rework and fewer strategy violations in briefs (measured by a simple checklist: ICP match, positioning match, perception addressed, revenue lever named). Guardrails = strategy file stays readable and maintained (no sprawling CLAUDE.md; strategy changes are committed in GitHub). Stop-loss = if the file isn’t referenced in the majority of briefs/reviews after two weeks, pause and fix workflow adoption before adding more content.
Readout: In the weekly meeting, don’t ask “did we like the output?” Ask two operator questions: (1) where did the strategy file force a better decision, and (2) where did it block progress because it was unclear or outdated? Then update the file.
Next test: Add the Perceptions and Positioning sections next, because those are the usual sources of inconsistency across channels. Different briefs often imply different stories. This catches it.
Trade-off: This will slow down shipping early. Strategy friction is the point. If speed is the only KPI, the system will be bypassed and Claude will revert to producing plausible nonsense on demand.
When this is wrong: If the business is genuinely in discovery mode (ICP unknown, product shifting weekly), locking decisions into a strategy file can create false certainty. In that case, the file should explicitly label assumptions as assumptions and time-box them—still written down, still versioned, just not treated as settled.
The kicker: the strategy doc didn’t fail—your workflow did
Kramer’s underlying critique lands because it’s familiar: teams create strategy documents but don’t use them. The result isn’t just misalignment. It’s wasted motion—campaigns that don’t compound because they’re not built on the same set of choices.
The interesting part of a /marketing-strategy skill in Claude isn’t “AI for strategy.” It’s making strategy unavoidable. Put the decisions in marketing-strategy.md. Reference it in everything. Track changes in GitHub. And keep the global instructions short enough to stay human-readable. The system doesn’t guarantee good judgment, but it does something most teams never operationalize: it makes forgetting harder than remembering.
