Operator-voice how-to guides and honest takes. Every post calls out where you're wasting Claude usage, time, or budget. Paired with what to do instead. No fluff, no listicles, no "10x your output." If a tactic doesn't earn its place, I cut it.
Generic is not a Claude problem. It is a prompt problem. The five specific patterns that produce vague answers, with the exact rewrite for each, plus the order to fix them in.
What to actually put in a Claude Project's Knowledge slot. The 7 files that matter, the 5 files that look useful and aren't, and the quarterly refresh ritual.
A complete worked Skill build (brand voice extractor) with the YAML frontmatter, the SKILL.md instructions, the install, and the mistake everyone makes on their first one.
Every "Claude prompt checker" tool in the SERP is actually a generator that rewrites your prompt for you. Here's the six-lever check a real checker uses, with a worked 1-out-of-6 prompt rewritten to 6-out-of-6.
The three containers Claude gives you, mapped to the work a real marketer does in a week. Skills for what is always true. Projects for what is true this quarter. Plain prompts for what is true once.
Market-level teardowns won't help you win a B2B account. The four prompts that produce intel a deal team can actually use, and the refuse-conditions that stop you shipping confident bullshit built on three reviews and a hunch.
Most Claude Projects rot in production. The 90-minute quarterly refresh that keeps them from becoming stale brand-shaped noise generators.
Every "ultimate Claude command" cheat sheet is fiction. The slash does nothing. The six structural levers that actually change the output, and the one place real commands exist.
How you talk to Claude shapes everything you get back. The mechanism, the limits, and why dry linear prompts produce dry linear output.
The diagnosis: adjective inputs produce adjective outputs. Three structural moves that change the result, and why a voice guide is for a person, not a brand.
A curated library is a refused library. Six categories of prompt I cut on sight, and why your stack should refuse them too.
Not the prompts I run during the work. The ones I run the day before so I walk in with hypotheses, not discovery questions.
Most "100 best prompts" listicles are useless. Here's the four-trait test for whether a prompt earns its place, and the ten that survive it.
Where each one earns its $20/month for a marketer specifically. The per-task verdict after 18 months running marketing on both.
Anthropic shipped Skills quietly. Most marketers haven't noticed yet. Here's what they are, when to use them, and four free marketing Skills to install today.
The exact four-section system prompt. The multi-brand pattern. The stale-Project problem. Ten minutes once, hours back every week.
Voice transfer from adjectives doesn't work. Voice transfer from examples does. Here's the exact prompt and the walkthrough.
Most AI content briefs produce 3,500-word generic posts. Here's the prompt setup that produces briefs that beat the SERP on something specific.
The four-hour competitive deck is procrastination. Here's the prompt that does the same job in thirty minutes, with positioning hypotheses and kill criteria.
Most "agent" setups are over-engineered. Claude Projects do 80% of the work. Here are the five that actually earn the setup time.
The five-test audit you should run monthly on every prompt in your stack, including the ones you wrote yourself.
A real week in the life. The work that actually pays back the $20/seat, and the work that quietly drains the calendar.