agentleFS
l-gevityl-gevity/l-gevity-skills/.claude/skills/architecture-as-code-javascript/SKILL.md

architecture-as-code-javascript

Skill88 starsChanged 3 days agoMITView on GitHub

JavaScript / TypeScript implementation of the `architecture-as-code` pattern. Per-subsystem `eslint.architecture.mjs` files merged into a single ESLint flat-config and enforced via `eslint-plugin-boundaries`. TRIGGER when: implementing or extending architecture-as-code in a JS/TS repo, debugging an `eslint-plugin-boundaries` rule, or adapting the assembler. SKIP for routine edits inside a governed subsystem. Reads in conjunction with `architecture-as-code` (the pattern, source of truth for schema, rule placement, anti-patterns, and audit checklist) — this skill defines only the JS-specific encoding, assembler code, and gotchas.

The file

We split this skill into the rules it gives an agent

2 rules. About 0 are common, 2 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Prove it red first: Add a throwaway file that imports the package from outside the adapter, confirm the lint fails on it with the declared …

    line 125
  2. Distinctive:

    When applying this implementation, emit:

    line 206

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Architecture-as-Code — JavaScript Implementation

Prerequisite. Read architecture-as-code first. The schema (§1), subsystems (§2), forbidden edges (§3), rule placement (§4), assembler concept (§5), and anti-patterns / audit (§6) are defined there and apply identically here. This file documents only what is JavaScript-specific.

1. File format

  • Filename: eslint.architecture.mjs. Use .mjs only; .js trips source-discovery walkers and ESLint's own config-loader.
  • ES module with export default { subsystems: [...], forbidden: [...] }, plus an optional externals: [...] for npm-package policy (§3). forbidden relates declared subsystems to each other and cannot name a package. components is read as a deprecated alias for subsystems (§2).
  • Pattern syntax: filesystem globs (<dir>/**).
  • Repo-root package.json must include "type": "module".
// eslint.architecture.mjs — example for a subsystem with internal layering
export default {
    subsystems: [
        { name: 'core-facade', pattern: 'packages/core/index.js', mode: 'file' },
        { name: 'core-tier1',  pattern: 'packages/core/tier1/**' },
        { name: 'core-tier3',  pattern: 'packages/core/tier3/**' },
        { name: 'core-other',  pattern: 'packages/core/**' },        // catch-all, last
    ],
    forbidden: [
        // Efferent — self-knowledge, lives in own file.
        { from: 'core-*', to: '*', except_to: ['core-*'],
          why: 'Core purity: no imports outside the core directory.' },
        // Internal layering — own-prefix only.
        { from: 'core-tier3', to: 'core-tier1',
          why: 'Tier 3 must go through tier 2.' },
    ],
};

Boundaries-plugin's parametric "not equal" syntax for cross-domain isolation is !{{from.captured.domain}} (double braces).

2. Assembler

eslint.config.js discovers every eslint.architecture.mjs deeper-first at lint startup, merges and expands them, and emits one eslint-plugin-boundaries block whose files glob equals the linted source set, with the file-existence and dependency rules both at error. Read references/assembler.md before creating or changing eslint.config.js, the assembler, or that block's files, severity, or ignores: it holds the code and why each line is shaped the way it is.

3. JavaScript-specific enforcement recipes

Confine a provider SDK to its adapter

The architecture-guidelines §10 handoff usually arrives as "only the adapter may talk to the provider". A provider SDK is an npm package, not an element, so forbidden cannot express it: that list only relates declared subsystems to each other. Declare package policy separately.

// eslint.architecture.mjs
export default {
    subsystems: [
        { name: 'email-adapter', pattern: 'api/src/email/**' },
        { name: 'api-other',     pattern: 'api/src/**' },   // catch-all, last
    ],
    forbidden: [ /* element-to-element edges */ ],
    externals: [
        { package: '@azure/communication-email', allow: ['email-adapter'],
          why: 'The provider SDK is confined to its adapter; callers use the port.' },
    ],
};

Assembler translation. Policies are evaluated in order and the last match wins, so deny the package everywhere first, then re-allow its owning elements:

const allExternals = archs.flatMap(m => m.default.externals ?? []);
const externalRules = allExternals.flatMap(x => [
    { from: expand('*'),
      disallow: { to: { module: { origin: 'external', source: x.package } } },
      message: x.why },
    { from: expand(x.allow),
      allow: { to: { module: { origin: 'external', source: x.package } } } },
]);

'boundaries/dependencies': ['error', {
    default: 'allow',
    checkAllOrigins: true,                       // REQUIRED — see below
    rules: [...rules, ...externalRules],
}],

Two traps make this silently wrong:

  • Without checkAllOrigins: true the policy never fires. The rule reports nothing, the lint is green, and the boundary does not exist. The flag is not an optimization.
  • checkAllOrigins: true subjects every package to the block's default. With default: 'disallow' that turns one SDK policy into a repo-wide import ban. Keep the blanket default: 'allow' and express prohibition as rules, or add an explicit allow-all policy first.

Register the rule at error. At warn this is a report, not a boundary.

Prove it red first. Add a throwaway file that imports the package from outside the adapter, confirm the lint fails on it with the declared message and line, delete the file, then confirm the adapter and its callers pass. A policy that has never failed has not been shown to run — the checkAllOrigins trap above produces exactly the same green output as a working rule. Keeping it proven afterward is gate-lifecycle.

Require literal dynamic-import paths

A computed import(expression) can bypass path resolution and therefore every subsystem boundary. Apply this rule to every production and test source block:

{
    files: ['**/*.{js,jsx,mjs,ts,tsx}'],
    rules: {
        'no-restricted-syntax': [
            'error',
            {
                selector: 'ImportExpression[source.type!="Literal"]',
                message:
                    'Dynamic import paths must be string literals so dependency boundaries remain enforceable.',
            },
        ],
    },
}

This permits import('./known-module.js') and rejects variables, concatenation, and template expressions. Keep the rule in the same flat config as the boundaries rules so a new source block cannot silently omit it.

Prevent production imports of test-only code

Declare test code as a narrower subsystem before the production catch-all, then forbid the production subsystem from importing it:

export default {
    subsystems: [
        { name: 'app-test', pattern: 'src/**/*.test.{js,jsx,ts,tsx}' },
        { name: 'app-test-support', pattern: 'test/**' },
        { name: 'app-prod', pattern: 'src/**' }, // catch-all, last
    ],
    forbidden: [
        {
            from: 'app-prod',
            to: 'app-test*',
            why: 'Production code must not import test-only code.',
        },
    ],
};

Tests may still import production subsystems. Adapt the globs to the repository, but retain the direction: production → test is forbidden.

Classify repository-root files

Root-level loose files — configs, entry scripts, generators — are the ones most often left unclassified, and the pattern's catch-all directive does not help here. A folder-mode ** at the root matches at the shallowest segment and claims files that specific elements already own, from any position in the list (see § 5). Declare the root's files with mode: 'file' instead:

// eslint.architecture.mjs (repository root)
export default {
    subsystems: [
        { name: 'repo-tooling-config', pattern: '*.config.js', mode: 'file' },
        { name: 'repo-tooling-entry',  pattern: 'server.js',   mode: 'file' },
        // No '**' catch-all at this level.
    ],
};

boundaries/no-unknown-files then names each root file still unclassified, so the list comes from the linter's output rather than from memory.

4. Output Contract

When applying this implementation, emit:

Scope:          <repo / package / subsystem path>
Decision:       Add eslint.architecture.mjs | Update assembler | Update ESLint config | Blocked
Generated config:<path, if any>
Rules changed:  <boundaries/dependencies element edges, externals package policy, or no-restricted-imports entries>
Verification:   <eslint command / assembler command / red-first proof / Not run + reason>
Next action:    <specific file edit, dependency install, or unresolved question>

5. JavaScript-specific gotchas

[!NOTE] Unresolved imports bypass enforcement. eslint-plugin-boundaries only enforces rules on imports it can resolve to a file path. Host-served absolute paths (e.g. SWA's /js/...) aren't resolved by default and pass silently. Fix: install eslint-import-resolver-alias and add it under settings['import/resolver'] so /js → packages/.../js resolves.

[!IMPORTANT] The rule block's files glob is a second, independent gate. A correct element registry, correct forbidden edges, and strict severities do nothing on a path outside files. A narrowed glob — ['packages/**/*.{js,ts,mjs}'], or an allowlist that grew one directory at a time — lets an entire new top-level directory into the repository with the boundaries lint reporting zero violations, because the rule never ran on it. Nothing in the output distinguishes that from a clean run. Keep the scope at ['**/*.{js,jsx,mjs,ts,tsx}'].

[!IMPORTANT] mode decides what a pattern matches, and the default is folder. In folder mode a pattern is tested against a file's path ancestors, so ** matches at the shallowest segment and captures files that deeper, more specific elements already own — regardless of its position in the list. A root-level { name: 'repo-unclassified', pattern: '**' } declared last, after several dozen specific elements, still wins, and yields a flood of misclassifications rather than the intended safety net. Use mode: 'file' for anything a shallower pattern would swallow.

[!NOTE] mode is deprecated in v7 and the suggested replacement does not cover this case. The deprecation warning recommends partialMatch: false; that does not classify a repository-root file — with a glob or with an exact filename, the file stays unknown. mode: 'file' is currently the only form that works. Accept the warning instead of chasing it.

[!NOTE] boundaries/external still works and is deprecated in v7. Unlike the mode deprecation above, this one has a working replacement: fold package policy into boundaries/dependencies with checkAllOrigins: true and the to.module.origin / to.module.source sub-selector (§3, Confine a provider SDK to its adapter). Do not run both rules over the same package — two verdicts on one edge, and the one that reports depends on rule order.

[!NOTE] Unmatched files bypass enforcement — silently. Files matching no element are invisible to boundaries/dependencies. End every constrained subsystem's subsystems list with a <dir>/** catch-all (pattern Directive 5), and set boundaries/no-unknown-files to error so an unmatched file is reported instead of ignored. Dependency rules are not a substitute: a file with no imports, or one loaded by a <script> tag, has no edge to judge.

[!NOTE] The plugin forbids; it never requires. An aspect coverage check — every subsystem in the governed set uses the mechanism — has no boundaries/dependencies form. dependency-cruiser's required rules can express it over the same registry; defect-shift-left places that check.

[!IMPORTANT] Flat config replaces a rule's options per rule id. Two blocks whose files overlap and that both set no-restricted-syntax or no-restricted-imports leave only the later block's entries active, and lint still passes. Merge the entries in the assembler.

[!NOTE] Emitted options are copied for every linted file. One selector per category can grow the boundaries options to megabytes and most of the lint time. Group element names per rule, measure uncached lint time after an assembler change, and guard the emitted size with a test.

[!NOTE] Facade-as-file pattern. JavaScript idiomatically exposes a facade as a single index/entry file. Use mode: 'file' plus an exact-path pattern (no glob) so the facade is matched alone.

[!NOTE] .js config breaks discovery. Many source-discovery walkers (the assembler's own, plus some lint plugins) treat .js as analyzable source. Naming the architecture file .js triggers self-reference and mis-classification. .mjs is required.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.