optbrew

한국어

A search-readiness audit skill for coding agents. optbrew checks a site for Google, Naver and AI answer engines, fixes the local files you approve, checks again under the same conditions, and writes a scored report built from replayable evidence.

• One skill for five profiles: SEO, NEO (Naver), AEO, GEO and LLMO.
• One call for a beginner: answer one set of start questions, then the agent runs the whole audit and hands you an HTML report.
• Five pinned audit guide bundles merged into one list of check families. An issue that several guides report is counted once.
• Every verdict in the report can be recomputed from the raw responses that were collected and hashed.
• No headline score when the evidence is thin: below 80% coverage the report shows the measured part and its bounds.
• Works with Codex, Claude Code, Cursor, OpenCode and Pi. Scoring and reports need no API account.

Why optbrew

Being found now means three different things: ranking in Google, showing in Naver, and being cited in AI answers. Each has its own guides, and the guides overlap and disagree. An agent asked to "improve SEO" usually picks one checklist, edits files, and reports success without proof.

optbrew gives the agent one workflow and one scoring rule for all five profiles. It counts what it measured, marks what it could not measure as unknown instead of passing it, and keeps a local fix separate from a production result. A high readiness score is not a promise of traffic or citations, and the report says so.

How it works

1. Ask. You give the URL. optbrew reads the start page, the sitemap and your setup (Chrome, a Google key, network access). Then it sends one message with the start questions: your guidance level, the goal, the questions people search for, the facts it read from the site, and the page limit. The questions, the guides and the report are in Korean. Reply with an answer for each number, or just "시작" (start) to take the defaults. If one thing blocks the audit, such as a network that is closed, you get one setup guide first. Up to three more guides appear as one-line offers; you open one by answering with its word, such as "키" (key).
2. Audit. One command collects the pages, renders up to 10 of them (one of each kind of page) in a headless Chrome of its own, measures speed and runs the code checks. It audits pages from the sitemap and from links on the start page, up to your limit: 50 pages by default, up to 200 if you raise it. The limit counts every request, so redirects and the robots and sitemap files use part of it. It follows a redirect to the www or https twin of your address and says so in the report. It stores every response with its hash.
3. Judge. The agent reads the page text that code cannot judge, such as claims, answer clarity and freshness. It cites the sentences it relied on. A second call of the same command checks that the cited text exists in the evidence, then scores.
4. Report. You get three report files with the same content: report.html (open this one; it is a single self-contained file), report.md and report.json. A fourth file, fixes.md, lists every open fix with how to make it and how to check it. The text is in Korean. Each problem has one fix with the number of pages it affects. A numbered list of steps tells you how to fill each blank, such as a missing Google key. Blanks are shown as unknown, never as a pass.
5. Fix and re-check (optional). The agent proposes changes to local files. Nothing is edited, deployed or submitted without your approval. After you deploy, the same collection runs again under the same conditions, so the before and after are comparable.

What the script measures by itself, and what still needs the guides and the agent's judgment:

Profile	Target	Measured by the script	Left to judgment
SEO	Google and web search	Fetch status, noindex in meta and X-Robots-Tag, titles, robots.txt for Googlebot, canonical, redirect chains, discovery through links and sitemap, JSON-LD syntax	Content quality and intent fit
NEO	Naver	robots.txt for Yeti, Open Graph, description, canonical, viewport, title repetition, blocked CSS or JS	Naver-specific content guidance
AEO, GEO	Answer engines and AI search	robots.txt for the AI search bots in ai-crawlers.md (training bots are never scored)	Answer structure, claims, authorship, freshness
LLMO	How models describe you	Whether name, address, phone and founding year agree with the facts you confirmed at the start and across pages, and whether the pages describe one business	Whether facts you did not confirm are true
All	Getting in touch (when a goal is conversion)	Whether an audited page offers a contact route: a tel:, sms: or mailto: link, a phone number shown as text, a KakaoTalk channel or open chat, Naver TalkTalk or Naver booking, or a link to an inquiry, consultation or booking page	Whether the route works and is easy to find
All	Speed and scripts	Whether the text still shows after JavaScript runs in Chrome. Field Core Web Vitals (LCP, INP, CLS) from Google's Chrome UX Report and PageSpeed Insights	Nothing. The Chrome lab values are shown as a run condition and are never scored

Your goals decide the target engines, and engine-specific checks run only for those. Google is always a target. ChatGPT, Claude and Perplexity are added together when a goal is AI answer citation, as in the default goals; a robots.txt that blocks their search bots lowers the score but does not block the audit. Naver is added when a goal names it, you give a Naver export, or the site is Korean (lang="ko" or an address ending in .kr). Whether the site is registered in Naver Search Advisor is not scored: without the verification meta tag or a Naver export, the report lists it under remaining work. The scoring rule is in scoring.md. The 80% coverage threshold is optbrew's own rule, not a search engine standard.

Install

Requires Node.js 22 or later. Python 3.10 or later is optional; without it, the internal Python audit utilities stay unrun. Pick one of the five methods. The examples install for Codex and Claude Code at user level; the agent IDs are codex, claude-code, cursor, opencode and pi.

In Windows PowerShell, if npx or optbrew reports that running scripts is disabled, type npx.cmd or optbrew.cmd instead.

1. npm

npm install -g optbrew@0.3.1
optbrew install --agents codex,claude-code --scope user --dry-run
optbrew install --agents codex,claude-code --scope user
optbrew doctor --agents codex,claude-code --scope user

The dry run prints the folders it will write. doctor checks the files and their hashes.

2. npx

Runs the same CLI without a global install:

npx optbrew@0.3.1 install --agents codex,claude-code --scope user
npx optbrew@0.3.1 doctor --agents codex,claude-code --scope user

3. PowerShell (Windows)

irm https://raw.githubusercontent.com/nowwcastle-sudo/optbrew/v0.3.1/install.ps1 | iex

The script is pinned to this release. It checks Node.js, shows the target folders, installs for Codex and Claude Code, and stops with an error if doctor finds a problem. To read it first, or to choose agents or a project folder:

irm https://raw.githubusercontent.com/nowwcastle-sudo/optbrew/v0.3.1/install.ps1 -OutFile install.ps1
powershell -ExecutionPolicy Bypass -File .\install.ps1 -Agents codex,claude-code,cursor

Add -Scope project -Project C:\path\to\site to install into one project.

4. Plugin

Claude Code:

claude plugin marketplace add nowwcastle-sudo/optbrew
claude plugin install optbrew@optbrew

Codex:

codex plugin marketplace add nowwcastle-sudo/optbrew
codex plugin add optbrew@optbrew

claude.ai, Claude Desktop and Cowork: open Customize > Plugins, choose Add > Add marketplace, enter nowwcastle-sudo/optbrew, then add optbrew.

Plugins follow this repository's main branch and update through the host's plugin commands or marketplace sync.

5. Ask your agent

Paste this into Codex, Claude Code or another coding agent:

Install the optbrew skill for me. Read https://raw.githubusercontent.com/nowwcastle-sudo/optbrew/v0.3.1/AGENT_INSTALL.md and follow it step by step. Ask me which agents and which scope to use, and show me the planned changes before you install.

AGENT_INSTALL.md gives the agent a command and a pass check for each step, and the cases where it must stop and ask you.

Installing from a source checkout, the Agent Skills CLI, Pi packages and Claude web uploads are covered in platform support.

Quick start

Start a new agent session and ask, for example:

Run optbrew on https://example.com. Goal: show up in Naver and get cited in AI answers.

optbrew then checks your setup and sends the start questions in one message. Answer each number, or reply "시작" (start) to accept the defaults. A missing Google key, Chrome or export file does not stop the audit. The blank is listed in the report with the steps to fill it.

The audit takes a few minutes. Everything it writes goes to optbrew-audit/<site>-<date> in your working folder: the evidence (every response with its hash), the three reports and fixes.md. Open report.html first. The same name with a number added (-2, -3) is used when the folder already exists. Private, internal and local addresses (such as localhost or 192.168.x.x) are refused, so audit the public address of the site.

Google API key (optional)

A free Google API key makes the real-visitor speed scores (LCP, INP, CLS) fill reliably. When optbrew offers it, answer "키" (key) to see how to get one. Then save it from your own terminal, never in the chat. In Codex and Claude Code cloud, the key guide shows where to save it instead.

optbrew key set google
optbrew key status
optbrew key remove google

set asks for the key with hidden input. status shows only whether a key exists and the first 8 characters of its hash. remove deletes it. If you did not install with npm, use npx optbrew@0.3.1 key set google, or answer "키" and copy the exact command from the guide. The skill folder has the same command too: node <skill folder>/scripts/key.mjs set google. The key is saved in .optbrew/credentials.json in your home folder, outside every project, and the environment variable OPTBREW_GOOGLE_API_KEY takes priority if it is set. With the Claude desktop app, use the key file (key set google): measured on 2026-10-18 on Windows, the plugin setting below did not reach a desktop-app session (bundled claude.exe 2.1.299), while terminal claude 2.1.302 received it; whether the app or the version makes the difference is not yet known. With the Claude Code plugin in terminal claude, you can instead open a terminal, run claude, and enter the key in /plugin, Installed, optbrew, Configure options (the desktop app's Code tab has no /plugin screen). Claude Code keeps it in your system credential store, a new session picks it up, and it comes before both only when it reaches the session. key status shows which source is used in google.sources. optbrew sends the key to Google only in a request header, and never writes it to a report, a log or the evidence.

Update

npm install -g optbrew@latest
optbrew update --scope user

With npx, run npx optbrew@latest update --scope user. optbrew check-updates --package optbrew tells you, without changing anything, whether npm has a newer release than the CLI you are running; optbrew doctor --json shows the release_version installed in your agent folders. npm update alone does not touch copies already placed in agent folders; the update command does.

Upgrading from 0.1.x

0.2.0 changes the evaluation version (0.4.0), the audit protocol (0.2.0) and the contract schema (2.0.0). So:

• A managed install of any 0.1.x release stops with migration_required, and optbrew update changes nothing. Follow Reinstall after a migration_required update; the steps delete only paths that a read-only check printed and never touch your audit evidence.
• A plugin install (Claude Code, Codex, claude.ai) follows main, so the plugin update commands of your host work as before.
• An output folder or evidence packet from a 0.1.x run is refused with one sentence: run the audit again in a new folder. Scores from different evaluation versions are not comparable, so compare reports not_comparable against a 0.1.x baseline. Run the baseline audit again with 0.2.0.

Upgrading from 0.2.x

Release 0.3.0 changes the evaluation version (0.5.0) and the audit protocol (0.3.0), because goals now choose checks. A managed 0.2.x install stops with migration_required in the same way; use the same reinstall steps. A 0.2.x output folder is refused, and compare reports not_comparable against a 0.2.x baseline, so run the baseline audit again.

APIs and cost

Nothing here requires a paid account. A direct Anthropic API call or an SE Ranking call needs your approval, a recorded budget and a stop condition first; a Claude subscription does not include API billing. The Google key above is your own, and the Google speed services it calls are free. Keys stay in your environment variables or the key file and never go into a report. Details: API boundaries.

Documentation

• Platform support: install paths per agent, checkout commands, reinstall steps
• Scoring, measurement and report evidence
• Changelog and release process

Contributing

Contributions are welcome: bug reports, false positives from a real site, guide updates, translations and code. See CONTRIBUTING.md.

License

optbrew's own code is MIT licensed. The bundled guides keep their own licenses, listed in third-party notices. The Naver bundle keeps an index of the official Search Advisor guides but not the 55 article bodies; follow the official links while online.