architecture-explorer
A Claude Code skill: an interactive, live architecture map of your repository (layers, API routes, tables, permissions, screens, services), every link traced to its file and line.
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/lemonadehighway--architecture-mapping)HTML
<a href="/listings/lemonadehighway--architecture-mapping"><img src="/api/listings/lemonadehighway--architecture-mapping/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.
You need [Claude Code](https://claude.com/claude-code) and Node.js 20 or later.
**As a plugin** (from Claude Code):
/plugin marketplace add LemonadeHighway/Architecture-Mapping /plugin install architecture-explorer@architecture-explorer
**As a plain skill:** copy `skills/architecture-explorer` into `~/.claude/skills/` (for all your projects) or into a project's `.claude/skills/`.
Then, in any repository, ask Claude something like *"show me the architecture of this project"* or *"open the architecture explorer"*. Claude starts the explorer, opens the map, and can set up a config for your project's own conventions.
**Without Claude:**
node skills/architecture-explorer/explorer/server.mjs --root /path/to/your/repo
node skills/architecture-explorer/explorer/cli.mjs --root /path/to/your/repo
The server prints its address (`http://127.0.0.1:5190` or the next free port). The CLI prints a summary, the coverage table and the findings; `--node <id>` and `--find <text>` answer questions from the terminal.
The explorer has no runtime dependencies. For TypeScript and JavaScript files it uses the TypeScript parser from the scanned project's `node_modules`; if the project has none, run `npm install` once in `skills/architecture-explorer/explorer` (it installs only `typescript`).
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 · 12 lines
- {
- "name": "architecture-explorer",
- "owner": { "name": "Zaid Zaatreh" },
- "plugins": [
- {
- "name": "architecture-explorer",
- "source": "./",
- "description": "An interactive, live architecture map of your repository, with every relationship traced to its source.",
- "version": "0.1.0"
- }
- ]
- }
README.md · README · 101 lines
- # Architecture Explorer
- A Claude Code skill that turns a repository into an interactive, live architecture map: layers, modules, API
- routes, database tables, permissions, screens, outside services and the business workflow, with every relationship
- traced to the file and line it comes from.
- 
- Built for people who are building a system and want to *see* it: what calls what, who may call which address, which
- tables a feature touches, where a new feature belongs, and what changed. It runs on your computer, reads the code,
- and never changes it.
- ## What you get
- - **Eleven views of one scan:** overview, software architecture, functional modules, code structure, module
- dependencies, frontend ↔ backend, API connections, database, business workflow, security & permissions, external
- integrations.
- - **Explore anything:** double-click a box to see what uses it, what's inside, and what it uses. Click a line to see
- the code behind it; every `file:line` opens the source.
- - **Honest about certainty:** each relationship is *verified* (a literal fact in the code), *inferred* (a name match
- or a convention) or *unresolved*. A Coverage tab compares raw text matches with what was understood and lists
- everything left over.
- - **Health findings:** addresses that change data with no authorization, sign-in-only writes, frontend calls that
- match no route, routes nobody calls, permissions nothing checks, tenant tables without row-level security, secrets
- in settings files. Each comes with its places and a ready prompt for Claude Code.
- - **Live:** it watches the repository and updates the open page when files change, with git's uncommitted changes
- marked.
- - **Search and Ask:** find files, routes, tables, permissions; ask "who calls /api/orders" or "which addresses are
- open" (index search, no AI, nothing leaves your machine).
- 
- ## What it understands
- | Area | Understood |
- |---|---|
- | C# / ASP.NET Core | Minimal APIs (`MapGet`…, route groups, helper methods that map routes), controllers with attribute routes, `[Authorize]` / `[AllowAnonymous]`, `RequireAuthorization`, `AddPolicy`, `IsInRole`, the fallback policy, DI registrations, handler and constructor injection, project references |
- | Data | `.sql` migrations replayed in order (tables, columns, foreign keys, indexes, triggers, row-level security), SQL in C# strings (Dapper, Npgsql, ADO.NET), EF Core `DbSet`s |
- | Frontend | TypeScript / JavaScript with `fetch`, `axios`, `api.get(...)` clients and `apiGet`-style helpers matched to server routes; imports with `tsconfig` path aliases; React Router, switch-based routers and Next.js `app/` / `pages/` screens |
- | Outside services | URLs in code, and well-known client packages (PostgreSQL, SQL Server, Redis, Stripe, Twilio, SendGrid, AWS, Azure, Auth0, Sentry, OpenAI, ...) |
- | Project conventions | Your own permission enum, route guards and permission checks, and your documented business process, through an optional `.architecture-explorer.json` ([reference](skills/architecture-explorer/references/configuration.md)) |
- Not modeled yet: EF Core migrations written in C#, JavaScript backends (Express, Next.js API routes), GraphQL, gRPC.
- C# is read by a C#-aware lexer, not compiled, so links between C# types are name matches and marked inferred.
- ## Install
- You need [Claude Code](https://claude.com/claude-code) and Node.js 20 or later.
- **As a plugin** (from Claude Code):
- ```
- /plugin marketplace add LemonadeHighway/Architecture-Mapping
- /plugin install architecture-explorer@architecture-explorer
- ```
- **As a plain skill:** copy `skills/architecture-explorer` into `~/.claude/skills/` (for all your projects) or into a
- project's `.claude/skills/`.
- Then, in any repository, ask Claude something like *"show me the architecture of this project"* or *"open the
- architecture explorer"*. Claude starts the explorer, opens the map, and can set up a config for your project's own
- conventions.
- **Without Claude:**
- ```bash
- node skills/architecture-explorer/explorer/server.mjs --root /path/to/your/repo
- ```
- ```bash
- node skills/architecture-explorer/explorer/cli.mjs --root /path/to/your/repo
- ```
- The server prints its address (`http://127.0.0.1:5190` or the next free port). The CLI prints a summary, the coverage
- table and the findings; `--node <id>` and `--find <text>` answer questions from the terminal.
- The explorer has no runtime dependencies. For TypeScript and JavaScript files it uses the TypeScript parser from the
- scanned project's `node_modules`; if the project has none, run `npm install` once in
- `skills/architecture-explorer/explorer` (it installs only `typescript`).
- ## Privacy and safety
- - Listens on 127.0.0.1 only and refuses requests for any other host name.
- - Read-only: it never writes to the repository; it runs only read-only git commands (`status`, `log`, `rev-parse`).
- - Only files the scan indexed can be opened from the page; secret-looking values in JSON settings are masked.
- - No network calls, no telemetry, no AI in the page.
- ## Develop
- ```bash
- cd skills/architecture-explorer/explorer
- npm install
- npm test
- ```
- The tests scan a small sample shop (`test/make-fixture.mjs`) end to end, check each scanner on small inputs, and
- start the real server to prove that a file change reaches the open page.
- ## License
- MIT, see [LICENSE](LICENSE).
Ratings and comments
Sign in to rate this skill.
Sign inFree to use. If it’s helped, pass it on.