10 claude code slash commands that run my one-person business

elisabeth hitz · july 27, 2026 · 4 min read

every list of claude code slash commands is written by a developer, about the built-in ones. /clear, /compact, useful, covered in the full guide. this list is different. these are ten commands i wrote myself, as markdown files, and they are the operating layer of a one-person business: 75+ live sites, 106 custom skills, no engineering team, 10+ years of enterprise sales before any of it.

the built-ins are what the tool ships with. these are what you build into it.

1. /founders-brief, the context loader

one file holds who i serve, how i talk, which words are banned, what i sell and at what price. every work session starts by loading it. the payoff is boring and enormous: i never re-explain my business, and the first draft of anything already sounds like me. if you build only one command, build this one.

2. /topic-radar, what to make next

takes any raw idea or news item and rules on it before i spend a minute building: is the window still open, do i have a real receipt, which audience does it serve, what's the non-generic angle. a go or no-go with a reason. it replaced the worst part of content creation, which was guessing whether a topic deserved my time.

3. /grounded-content-generator, drafts from real data

generates content grounded in my own performance data. it loads the current winning patterns and my top recent posts before writing a word, so every draft is modeled on what already worked instead of vibes. paired with a fact gate, it's the reason i can publish daily without publishing nonsense.

4. /tiktok-hook-engineer, the first three seconds

engineers hooks against hard constraints: keyword in the first two seconds, seven to fourteen words, a fixed set of hook archetypes, maximum two variants so i don't drown in options. constraints are the whole value. "write me a hook" gets slop. a spec gets a hook.

5. /retention-engineer, the teardown

reads a finished script and returns timecoded fixes: where the hook leaks, where the pattern interrupt is missing, where the mid-video dead zone will kill completion. a brutal editor that never flatters me. every script i film goes through it first.

6. /content-deployer, one asset, every platform

takes one piece of content and maps it into the right native formats per platform, in posting order. what used to be an afternoon of reformatting is one command and a review pass.

7. /em-dash-sweep, voice enforcement

em dashes are the loudest AI tell in writing, so they're banned from my brand. this command hunts them across my pages and swaps each one for the punctuation a person would actually use: comma, period, or parens, depending on context. a writing rule i used to enforce from memory is now a program. (why em dashes give AI away, and the other tells, here.)

8. /standing-rules-audit, the rule keeper

my brand has locked rules: phrases that never appear, punctuation that never ships, claims that were banned by name. this command audits every published page against the whole rulebook at once. standards you can run beat standards you remember.

9. /weekly-revenue-audit, the money truth

once a week, one command pulls the actual numbers and reports what sold, what didn't, and what changed. not a dashboard i have to remember to open. a ritual that runs.

10. /kit-attribution-verify, which content actually converts

probes every email capture end to end: it submits a test signup and confirms the right source tag and sequence actually attach inside Kit, then cleans up after itself. that attribution is how i know which reel, which tool, which post actually produced a subscriber. without it, content strategy is astrology.

the pattern worth stealing

look at the list again. not one of these is a clever prompt. each one is a job description: context, research, drafting, hooks, retention, distribution, voice, rules, revenue, attribution. the business is decomposed into named, repeatable programs that happen to be markdown files.

that's the thesis this whole site runs on: files not chats, skills not prompts. a chat is rented. a file is owned.

if you want to build your own set, the 5-minute tutorial gets you your first command today, and the toolkit is 27 of mine for the business side, with the brand voice skill running free on the page before you pay anything.

go deeper

source: the ten live skill files in ~/.claude/skills this post describes; every description checked against the actual SKILL.md before publish, and the skill count verified against the live folder, july 27 2026. written by elisabeth hitz, certified in anthropic's ai fluency program (framework & foundations, and ai capabilities & limitations), plus claude 101 and claude cowork.