You Don't Have a Reputation Problem. You Have a Reading Problem.

You do not have a reputation problem. You have a reading problem.

Marcus runs a twelve-person auto detailing shop in Austin. Cars are his craft. Reviews are his oxygen. A five-star Google review brings him a walk-in. A bad TikTok clip of a rushed job can cost him a month of bookings before he even sees it.

So every morning, before coffee, Marcus opens five apps. Google Reviews. Yelp. Instagram comments. The Austin car subreddit. TikTok, because someone is always filming their detail. He reads, he screenshots, he replies where he can. Some mornings it takes forty minutes. Some mornings he misses something that mattered, because a customer complained on a platform he did not think to check.

Marcus does not have a shortage of customers talking about him. He has a shortage of ears.

What actually changed this week

On 16 August, a builder named Corey Haines shipped a small, free tool into the open-source Claude Skills ecosystem: a skill called social-fetch. An AI agent skill is a short instruction file, nothing more, that teaches an agent like Claude Code, Codex, or Cursor how to do one job well. You install it once. After that, the agent reaches for it automatically whenever the task fits, the same way you reach for a familiar tool without being told to.

Social-fetch's one job: give it any social post URL, from X, LinkedIn, Instagram, TikTok, Reddit, Bluesky, Threads, Mastodon, or Hacker News, and it hands back the same clean, structured answer every time: who posted it, what they said, how many people engaged, and the replies underneath. One format, ten platforms. The agent does not care that Instagram and Reddit are built completely differently. Neither do you, once it is set up.

Within a day, the post announcing it had gathered more than a thousand bookmarks. That is not proof of importance on its own, and Notma will not pretend one tweet is a market signal. But it is the fastest-spreading example this month of a pattern worth understanding on its own merits: agents that can read the open web on your behalf, in a structured way, are becoming a one-line install instead of a custom engineering project.

Isn't this just a fancier scraper?

Fair question. Scrapers have existed for years. Two things are different here.

First, the skill tries the cheapest, most reliable path first. Reddit, Bluesky, Mastodon, and Hacker News have open public APIs, so it reads those directly, free, every time. Platforms that lock content behind logins, like Instagram, TikTok, and specific LinkedIn posts, get a browser-based fallback, and only fall through to a paid provider if you have chosen to configure one. You are never paying for what you did not need to pay for.

Second, and this is the part that matters for a business owner rather than a developer: the output is always the same shape. Author, text, timestamp, engagement counts, replies. That means once it is installed, you do not learn a new tool for each platform. You ask your agent one kind of question, in plain English, and it goes and gets the answer no matter where the conversation is happening.

What is an AI agent skill, and how do I actually install one?

A skill is a plain markdown file with two parts. At the top, a short header tells the agent what the skill does and when to use it. Below that, the body holds the exact steps: what to check first, what to try next, what a finished answer looks like. The agent reads the header the way you would skim a job posting, decides whether the task in front of it matches, and only then loads the full instructions.

That is the whole mechanism. No new infrastructure, no monthly subscription bolted onto your existing AI tool, no vendor lock-in. If you already use Claude Code, Codex, or Cursor, you add the skill once, the same way you would add a browser extension, and it sits there ready.

The general pattern looks like this in Claude Code:

1. Add the skill's repository as a source: /plugin marketplace add coreyhaines31/makerskills
2. Install it: /plugin install makerskills@coreyhaines31
3. Or grab just the one skill you want with a single command: npx skills add coreyhaines31/makerskills
4. Ask your agent, in plain language: "Pull the last thirty posts mentioning [your business name] on Reddit and X, and tell me which ones need a reply today."

No API keys required to get started. No cloud account. It runs locally, inside the agent you already have open.

The old way, the expensive way, and this way

Set this up in one afternoon

1. Install the skill into whichever agent you already use for work, following the three-line install above.
2. Give it a starting list of what to watch: your business name, your owner's name if customers tag you personally, and two or three competitors worth tracking.
3. Ask for a same-day sweep. Marcus's actual first prompt: "Search Reddit, X, and Instagram for mentions of [shop name] from the last 48 hours, fetch each post with social-fetch, and summarise which ones are complaints, which are praise, and which need a reply."
4. Turn it into a habit, not a one-off. Ask your agent to run the same sweep every morning and hand you three lines: what people are saying, what needs a response today, and anything that looks like it could spread.
5. Reply like a human, fast. The skill gets you the read. You still write the reply. Speed is the whole win here, not automation of the relationship.

Total setup time for Marcus, start to finish: nineteen minutes, most of it spent deciding which two competitors to track.

What you have after this

A week in, Marcus stops opening five apps every morning. He opens one message from his agent instead.

A month in, he has caught two negative comments on Reddit before they turned into a pile-on, and replied within the hour both times, which is the single biggest lever for how a bad review resolves.

Six months in, he has a rough, honest record of what customers actually think, across every platform, not just the ones he happened to check. He uses it to fix the one thing detailers usually get wrong quietly: how long a job actually takes versus what was promised.

Same shop. Same twelve people. A completely different relationship with what is being said about them.