@cef-ai/eslint-plugin

Editor-time lint rules for CEF agents. Each agent rule mirrors a check that cef build performs, so warnings in your editor agree with build failures in CI.

The plugin also exports one architectural-boundary rule, no-web3-imports, which the SDK monorepo applies to its own source to keep chain/web3 imports confined to @cef-ai/account. It is not part of the agent build and is deliberately excluded from recommended — see Architectural rules below.

Install

pnpm add -D @cef-ai/eslint-plugin

Configure

The plugin ships a recommended config that turns every agent rule on (no-web3-imports is excluded — it is an SDK-internal rule, see below). The three "declared alias / type" rules accept options carrying the declared sets so the rule can flag undeclared usage without a type-checker.

// .eslintrc.cjs
module.exports = {
  plugins: ["@cef-ai"],
  extends: ["plugin:@cef-ai/recommended"],
  rules: {
    "@cef-ai/cubby-declared-alias": ["error", { aliases: ["history"] }],
    "@cef-ai/model-declared-alias": ["error", { aliases: ["llm"] }],
    "@cef-ai/publish-declared-type": [
      "error",
      { knownEventTypes: ["assistant_message"] },
    ],
  },
};

cef typegen is expected to emit a snippet wiring these option arrays from cef.config.ts (cubbies[].alias, models keys) and from the project's KnownEventTypes declarations / declared publishes. Until typegen drives this end-to-end, hand-write the arrays to match your config.

Agent rules (in recommended)

These are the rules cef build enforces. The recommended config turns all of them on.

| Rule | What it flags | | --- | --- | | @cef-ai/on-event-literal | @OnEvent(...) must take a string literal, never a variable or expression. | | @cef-ai/engagement-id-literal | An engagement id argument must be a string literal, never a variable or expression. | | @cef-ai/publish-declared-type | ctx.publish(t, ...) first argument must be a string literal AND a known event type. | | @cef-ai/cubby-declared-alias | ctx.cubby("alias") must use a string literal AND a declared cubby alias. | | @cef-ai/model-declared-alias | ctx.models.X must reference a declared model alias. | | @cef-ai/no-banned-imports | Imports must not target Node built-ins (fs, net, http, child_process, ...) or unsafe sync DB / network npm packages. |

cef build enforces the same rules and aborts the build on the first violation; this plugin gives you the same feedback at edit time without waiting for a build cycle.

Architectural rules (not in recommended)

| Rule | What it flags | | --- | --- | | @cef-ai/no-web3-imports | Chain/web3 imports (@polkadot/*, @cere-ddc-sdk/*, cere-ddc-sdk-js) outside @cef-ai/account. |

no-web3-imports enforces an SDK-monorepo boundary, not an agent-build check, so it is not in recommended and cef build does not run it. The monorepo applies it repo-wide in its own root .eslintrc.json and turns it off for @cef-ai/account (the one package where chain types belong); the optional allow list is an escape hatch for intentional exceptions. If you are fencing your own packages off from web3, enable it explicitly:

// .eslintrc.cjs (repo-internal, not for agents)
rules: {
  "@cef-ai/no-web3-imports": ["error", { allow: ["@polkadot/", "cere-ddc-sdk-js"] }],
}

Companion packages

• @cef-ai/agent-sdk — the decorators and Context shape these rules check against.
• @cef-ai/cli — cef build, where the same checks run as part of the bundle pipeline; full banned-import list lives in the agent-sdk spec.

References

• Agent SDK spec: ../../../company-memory-bank/specs/platform/03-components/agent-sdk.md §7 (lints + banned imports).

The ../company-memory-bank/... path points at an internal sibling repo and is not browsable from a fresh clone.