aispy
Safety scan
A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.
Badge for your README
Markdown
[](/listings/stevensmith-design--aispy)HTML
<a href="/listings/stevensmith-design--aispy"><img src="/api/listings/stevensmith-design--aispy/badge.svg" alt="findsafeskills safety rating" /></a>Install
Install instructions from the README
As written by the repo's author. findsafeskills has not verified these steps.
**From a clone or download** (nothing to install into Claude Code):
1. **Get AIspy:** `git clone https://github.com/stevensmith-design/aispy.git`, or download the ZIP from GitHub and rename the unzipped folder (`aispy-main`) to `aispy`. Open Claude Code in the folder that contains `aispy` (your project, for example), or connect the `aispy` folder to a session alongside your project. 2. **Ask Claude:** "Set up AIspy for my first review". It installs the screenshot helper and tests a scripted browser; approve the install and browser commands when it asks. 3. **Ask for a review:** "Use the AIspy kit in ./aispy to review https://example.com. Start separate agents for the lenses and checks." With the folder connected separately, say "the AIspy kit in the connected aispy folder".
**Or install it as a plugin:**
1. **Get AIspy:** `/plugin marketplace add stevensmith-design/aispy`, then `/plugin install aispy@aispy`. From a local copy: `/plugin marketplace add ./aispy`, then the same install command. 2. **Ask Claude:** "Set up AIspy for my first review". It offers to install the screenshot helper and test the browser; say yes and approve the commands. 3. **Ask for a review:** "Use the AIspy kit to review https://example.com. Start separate agents for the lenses and checks."
Each review starts with a setup check (agents, the browser reaching your site, screenshots, the kit) and stops if anything fails. The manual commands are under [Install](#install).
クイックスタート(日本語)
**クローンまたはダウンロードして使う**(Claude Code へのインストールは不要)
1. **AIspy を入手する:** `git clone https://github.com/stevensmith-design/aispy.git` を実行するか、 GitHub から ZIP をダウンロードし、展開したフォルダ(`aispy-main`)の名前を `aispy` に変更します。 `aispy` が入っているフォルダ(作業中のプロジェクトなど)で Claude Code を開くか、`aispy` フォルダをプロジェクトと一緒にセッションに接続します。 2. **Claude に依頼する:**「AIspy の初回レビューの準備をして」。スクリーンショット用のツールを インストールし、ブラウザの自動操作をテストします。インストールやブラウザのコマンドの許可を 求められたら承認してください。 3. **レビューを依頼する:**「./aispy の AIspy キットで https://example.com をレビューして。 レビュー観点とチェックは別々のエージェントで実行して。」フォルダを別に接続している場合は 「接続した aispy フォルダの AIspy キットで」と伝えます。
**プラグインとしてインストールして使う**
1. **AIspy を入手する:** `/plugin marketplace add stevensmith-design/aispy` を実行したあと、 `/plugin install aispy@aispy` を実行します。ローカルのコピーから入れる場合は `/plugin marketplace add ./aispy` のあと、同じインストールコマンドを実行します。 2. **Claude に依頼する:**「AIspy の初回レビューの準備をして」。スクリーンショット用のツールの インストールとブラウザのテストを提案してくるので、了承してコマンドを承認してください。 3. **レビューを依頼する:**「AIspy キットで https://example.com をレビューして。レビュー観点と チェックは別々のエージェントで実行して。」
レビューは毎回、セットアップチェック(エージェント、サイトへのブラウザ接続、スクリーンショット、 キット)から始まり、どれかが失敗するとそこで止まります。手動で実行するコマンドは [Install](#install) にあります。
What was scanned
The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.
.claude-plugin/marketplace.json · Manifest · 13 lines
- {
- "name": "aispy",
- "owner": {
- "name": "Steven Smith"
- },
- "plugins": [
- {
- "name": "aispy",
- "source": "./",
- "description": "UX/UI/IA review of a live website or app, with an optional market-localisation layer, styled in the product's own design system."
- }
- ]
- }
README.md · README · 276 lines
- # AIspy
- 
- **A full UX, UI and IA review from your AI crew.**
- Made by Steven Smith. More on AIspy and my other work: [stevensmith.dev/aispy](https://stevensmith.dev/aispy?utm_source=github&utm_campaign=aispy)
- | Review report | IA analysis | Mock |
- |---|---|---|
- |  |  |  |
- <sub>Real output from a full AIspy run on Brolga, a fictional running-shoe brand selling in Japan. The standard every run is held to is [`examples/mareo/`](examples/mareo/), the reference review recreated on a fictional store. More under [Examples](#examples).</sub>
- For product managers, founders, developers and marketers without a senior designer to hand, and for designers and agent-led teams who want a full review to start from.
- AIspy is a skill pack for Claude Code that runs a professional UX/UI/IA review with an optional market-localisation layer. Give it a product; it captures the pages, extracts the design system, researches the market and benchmarks, runs three independent review lenses, and
- delivers:
- - a **review report**: findings grouped by cause, each with annotated evidence and a site that does it well;
- - an **IA analysis**: navigation, page layout by type, heading hierarchy, task paths, label audit and a proposed structure;
- - a **current-vs-proposed mock** of the key screen, built in the product's own design system;
- - an **index page** that ties the three together.
- All four are styled in the reviewed product's own design system. AIspy replicates a reference
- review made by a senior designer, and was tested against it (November 2026): its research and
- findings at least match the reference, and its documents follow the same form.
- It supports a reviewer; it doesn't replace one. It stops for your review after the report,
- again after the mock, and before packaging.
- ## Quick start
- **From a clone or download** (nothing to install into Claude Code):
- 1. **Get AIspy:** `git clone https://github.com/stevensmith-design/aispy.git`, or download the ZIP
- from GitHub and rename the unzipped folder (`aispy-main`) to `aispy`. Open Claude Code in the
- folder that contains `aispy` (your project, for example), or connect the `aispy` folder to a
- session alongside your project.
- 2. **Ask Claude:** "Set up AIspy for my first review". It installs the screenshot helper and tests
- a scripted browser; approve the install and browser commands when it asks.
- 3. **Ask for a review:** "Use the AIspy kit in ./aispy to review https://example.com. Start
- separate agents for the lenses and checks." With the folder connected separately, say "the
- AIspy kit in the connected aispy folder".
- **Or install it as a plugin:**
- 1. **Get AIspy:** `/plugin marketplace add stevensmith-design/aispy`, then
- `/plugin install aispy@aispy`. From a local copy: `/plugin marketplace add ./aispy`, then the
- same install command.
- 2. **Ask Claude:** "Set up AIspy for my first review". It offers to install the screenshot helper
- and test the browser; say yes and approve the commands.
- 3. **Ask for a review:** "Use the AIspy kit to review https://example.com. Start separate agents
- for the lenses and checks."
- Each review starts with a setup check (agents, the browser reaching your site, screenshots, the
- kit) and stops if anything fails. The manual commands are under [Install](#install).
- ### クイックスタート(日本語)
- **クローンまたはダウンロードして使う**(Claude Code へのインストールは不要)
- 1. **AIspy を入手する:** `git clone https://github.com/stevensmith-design/aispy.git` を実行するか、
- GitHub から ZIP をダウンロードし、展開したフォルダ(`aispy-main`)の名前を `aispy` に変更します。
- `aispy` が入っているフォルダ(作業中のプロジェクトなど)で Claude Code を開くか、`aispy`
- フォルダをプロジェクトと一緒にセッションに接続します。
- 2. **Claude に依頼する:**「AIspy の初回レビューの準備をして」。スクリーンショット用のツールを
- インストールし、ブラウザの自動操作をテストします。インストールやブラウザのコマンドの許可を
- 求められたら承認してください。
- 3. **レビューを依頼する:**「./aispy の AIspy キットで https://example.com をレビューして。
- レビュー観点とチェックは別々のエージェントで実行して。」フォルダを別に接続している場合は
- 「接続した aispy フォルダの AIspy キットで」と伝えます。
- **プラグインとしてインストールして使う**
- 1. **AIspy を入手する:** `/plugin marketplace add stevensmith-design/aispy` を実行したあと、
- `/plugin install aispy@aispy` を実行します。ローカルのコピーから入れる場合は
- `/plugin marketplace add ./aispy` のあと、同じインストールコマンドを実行します。
- 2. **Claude に依頼する:**「AIspy の初回レビューの準備をして」。スクリーンショット用のツールの
- インストールとブラウザのテストを提案してくるので、了承してコマンドを承認してください。
- 3. **レビューを依頼する:**「AIspy キットで https://example.com をレビューして。レビュー観点と
- チェックは別々のエージェントで実行して。」
- レビューは毎回、セットアップチェック(エージェント、サイトへのブラウザ接続、スクリーンショット、
- キット)から始まり、どれかが失敗するとそこで止まります。手動で実行するコマンドは
- [Install](#install) にあります。
- ## Examples
- Two example packages. Each document is one self-contained HTML file; open `index.html` in a
- browser to start.
- | | [`examples/mareo/`](examples/mareo/) | [`examples/brolga/`](examples/brolga/) |
- |---|---|---|
- | What it is | The review AIspy was built and tested against, recreated on a fictional Danish sofa brand selling in Japan. It isn't a kit run: it shows what the kit aims at. | A full AIspy run on a fictional Australian running-shoe brand selling in Japan, reviewed once by a person. |
- | Scope | One product page at 390 px, with the header, menu drawer and search. Sections that need other pages are kept and labelled "out of scope in this example". | The top page, all shoes, three product pages and the size guide; phone first, desktop where a finding differs. |
- | What it shows | The standard the skill matches by default: document form, density of argument, how evidence is marked, and the mock. Also the store's product page before and after. | What a complete multi-page package looks like, from intake to the finished documents. |
- | Documents | Index, localisation review (13 findings), IA analysis, mock, mock spec | Index, review report (18 findings), IA analysis, mock, mock spec |
- Both pass the pack's checks (`scripts/checks/`). The examples' stores and products are made up;
- real Japanese sites appear only as benchmarks.
- ## Install
- Requires [Claude Code](https://code.claude.com), in the terminal or the desktop app's Code tab,
- Node 22.23.5 or newer, and Python 3 (for the design-system validator, `validate-tokens.py`).
- AIspy installs into Claude Code as a skill pack. Either:
- **With Claude Code's `/plugin` command:**
- ```
- /plugin marketplace add stevensmith-design/aispy
- /plugin install aispy@aispy
- ```
- **Or from a clone**, starting Claude Code with it loaded:
- ```
- git clone https://github.com/stevensmith-design/aispy.git
- claude --plugin-dir ./aispy
- ```
- To keep it with a project, put the folder inside it and start Claude Code with `--plugin-dir` and
- that path. Keep the folder whole: the skills find their scripts two folders up.
- Either way, eight skills become available: `product-review` and `review-scoping`, plus the six
- design-kit skills they call (`ui-design-system`, `ui-design-review`, `ia-review`, `design-refine`,
- `ui-design-principles`, `distinctiveness-review`). The design kit is bundled. To confirm AIspy is
- loaded, run `/plugin` (or `claude --plugin-dir ./aispy plugin details aispy`) and look for the
- eight skills; tested, the bare names resolve to AIspy's copies even when other copies of the kit
- are installed.
- **Before the first review**, set up what the setup check needs, or it stops:
- 1. **Screenshots:** install the capture helper (Playwright), which makes the evidence crops. Run
- it in the pack's `scripts/capture/` folder:
- - **From a clone**, in the folder where you ran `git clone`:
- ```
- cd aispy/scripts/capture && npm install && npx playwright install chromium
- ```
- - **With `/plugin` from a local copy** (`/plugin marketplace add ./aispy`): as from a clone. The
- plugin runs from that folder.
- - **With `/plugin` from GitHub**, in any folder. Claude Code keeps the plugin in a folder named
- after the installed version; this picks the newest installed version. Run it again after each
- update:
- ```
- cd "$(ls -d ~/.claude/plugins/cache/aispy/aispy/*/ | sort -V | tail -1)scripts/capture" && npm install && npx playwright install chromium
- ```
- 2. **A browser the session can drive**, such as Claude in Chrome or Playwright, with its commands
- allowed in your permissions so it can open the site.
- 3. **Agents:** say "start separate agents for the lenses and checks" in your request.
- ## How to start
- The only thing it needs is the product: a URL, a store link, or screens.
- ```
- Review https://example.com and propose improvements. Start separate agents for the lenses and checks.
- ```
- ```
- Here's the brief: <paste>. My direction: focus on mobile checkout, and the store is
- launching in Japan, so include a localisation review. Report in English.
- ```
- ```
- Extract the brand and design system from https://example.com first; we'll review afterwards.
- ```
- It takes a quick look, then asks one batch of questions with a proposed default for each, so "go"
- is enough. Before reviewing, it runs a setup check (agents, browser, screenshots, kit) and stops if
- anything fails.
- ### 始め方(日本語)
- 必要なのはプロダクトだけです。URL、ストアのリンク、または画面のどれかを渡してください。
- ```
- https://example.com をレビューして、改善案を出して。レビュー観点とチェックは別々のエージェントで実行して。
- ```
- ```
- ブリーフはこちら:<貼り付け>。方針:モバイルの購入手続きを重点的に見てほしい。日本市場でローンチ
- するので、ローカライズのレビューも含めて。レポートは日本語で。
- ```
- ```
- まず https://example.com からブランドとデザインシステムを抽出して。レビューはそのあとで。
- ```
- まずサイトをざっと確認し、質問をまとめて一度に送ります。各質問には既定の案が付いているので、
- 「進めて」と返すだけで構いません。レビューの前にセットアップチェック(エージェント、ブラウザ、
- スクリーンショット、キット)を行い、どれかが失敗するとそこで止まります。
- ## How to use
- [docs/how-to-use.md](docs/how-to-use.md) covers it in full. In short:
- - **Live site:** the URL is enough. Better, also save each page in scope as HTML ("Save page as",
- complete) and connect the folder.
- - **Staging or local site:** give the served build or a reachable URL, never the source project
- (the source holds the design intent and biases the review).
- - **Logged-in areas:** you sign in yourself in the browser; the pack never handles credentials.
- - **Native app:** a simulator or emulator the session can see and operate, or exported screens per
- flow with the sweep done by hand. Tap areas are less exact than on the web.
- - **Say in the request:** review type (general, or plus localisation), priority device, pages or
- flows, language, and "start separate agents for the lenses and checks".
- ## What to expect
- Three stops, one review round from you at each:
- 1. **Review report and IA analysis.** The deliverables step takes about 40 minutes from a finished review.
- 2. **Mock spec, then the mock.** About 25 minutes, including the spec and checks.
- 3. **Packaging:** the index page and the final package.
- Times are from the test runs. Everything is written to `review/<product>-<date>/` in the folder you
- connected, with a run log.
- **Accessibility:** it checks contrast, text size, tap targets, alt text, screen-reader strings,
- headings and motion. It is not a full WCAG audit, and it doesn't do a screen-reader or
- keyboard-only walk unless you ask.
- ## Contents
- ```
- .claude-plugin/ plugin and marketplace manifests
- skills/
- product-review/ SKILL.md (the workflow), references/ (checklist, review questions,
- deliverable outlines), templates/ (the four documents; see templates/README.md)
- review-scoping/ optional: inventory a big site and pick pages by type
- design-kit/ the six design-kit skills the pack calls (see design-kit/PROVENANCE.md)
- scripts/ optional helpers, not required steps
- capture/ capture.mjs, extract-in-page.js, sweep-in-page.js, annotate.mjs (evidence crops)
- (tests: node --test scripts/capture/test/)
- inventory/ inventory.mjs (sitemap, page types, broken URLs)
- build/ theme-from-tokens.mjs, assemble.mjs, layout-bars.mjs
- checks/ check-deliverable.mjs, check-report.mjs, check-mock.mjs, forbid.mjs
- (tests: node --test scripts/checks/test/)
- examples/mareo/ the reference review recreated on a fictional store (default standard)
- examples/brolga/ a complete multi-page example run
- docs/how-to-use.md inputs by product type, and what to expect
- ```
- ## How AIspy is designed
- Consistent results come from defining the deliverables, not from leaving them to the agent.
- 1. **Goals:** what each deliverable is for and who reads it.
- 2. **Data required:** what evidence, research, benchmarks and extraction each one needs.
- 3. **How it's put together:** the required contents, structure and layout of the report, IA
- analysis, mock and index, as rules, templates and examples. This is the baseline; an agent may
- add something new with a stated reason, but never starts from a blank page.
- 4. **Methods:** how to capture, extract, research, benchmark and review. This is where the agent's
- judgement works: what it finds, how serious it is, and what to propose.
- ## How it improves
- Through use, measured against the reference review. Treat any report that something is missing or
- worse as a gap in replicating the reference: check what the reference does, then repair the
- existing instruction or reference file. When a real run misses something, add one line to
- `skills/product-review/references/checklist.md`. A new requirement or step goes in only for a
- demonstrated gap, or when a run shows it beats the reference on quality or cost.
- ## Status
- 0.8.7. See [CHANGELOG.md](CHANGELOG.md) for the history and the
- [open issues](https://github.com/stevensmith-design/aispy/issues) for what's next.
- ## Licence
- Code, skills and templates: [MIT](LICENSE). The bundled design kit keeps its own MIT licence and
- notice (`skills/design-kit/LICENSE`, `NOTICE`). The example packages are
- CC BY 4.0: [`examples/mareo/`](examples/mareo/LICENSE.md) and
- [`examples/brolga/`](examples/brolga/LICENSE.md). The Barlow Condensed font embedded in
- Brolga's documents is under the SIL Open Font License 1.1, not CC BY (see its `LICENSE.md`).
- The AIspy name, logo and character artwork (including the image at the top of this page) are
- not covered by these licences.
Ratings and comments
Sign in to rate this skill.
Sign inFree to use. If it’s helped, pass it on.