ts-dev

name	ts-dev
description	Use when writing new TypeScript code, fixing bugs, refactoring, or adding tests — user says "implement", "fix", "refactor", "add tests", or is editing .ts, .tsx, tsconfig.json, or package.json files. Does NOT apply to reading or discussing existing code without changes.
slash-command	false

TypeScript Development

Quick Reference

Category	Rule	How to Apply
Type Safety	Never use any	Use unknown with type guards; any defeats the compiler
	Avoid type assertions	as Type suppresses errors without fixing them
	Avoid non-null assertions	! crashes on edge cases; use ?. and ?? instead
	Strict mode	strict: true in tsconfig.json is non-negotiable
	Discriminated unions	Model state explicitly; no boolean flags
Async Patterns	Always await	Unhandled rejections crash Node.js silently
	No await in loops	Use Promise.all for parallel; sequential only when required
	Partial failures	Promise.allSettled when one failure shouldn't abort all
Error Handling	Type-check caught errors	catch (e) is unknown in strict mode — narrow before use
	Never swallow errors	Log or rethrow; empty catch blocks hide bugs
	Result types	Use discriminated unions for expected failures
Testing	Prefer real implementations	Mocks drift from production; real integrations catch real bugs
	Test behavior not impl	Don't test internal state; test observable outcomes
	Type-level tests	Use expectType<T>() for type assertions
Code Quality	const over let	Prevents accidental reassignment; never var
	readonly where possible	Communicates intent; prevents mutation bugs
	Template literals	Clearer than concatenation; no escaping issues

Rule Priority Decision Flow

flowchart TD
    Writing_code((Writing code))
    Type_safety_violation_{Type safety violation?}
    Apply_Type_Safety_rules[Apply Type Safety rules]
    Async_issue_{Async correctness issue?}
    Apply_Async_rules[Apply Async rules]
    Error_handling_gap_{Error handling gap?}
    Apply_Error_Handling_rules[Apply Error Handling rules]
    Apply_Code_Quality_rules[Apply Code Quality rules]
    Code_complete((Code complete))
    Writing_code --> Type_safety_violation_
    Type_safety_violation_ -->|"yes (NEVER compromise)"| Apply_Type_Safety_rules
    Type_safety_violation_ -->|no| Async_issue_
    Apply_Type_Safety_rules --> Code_complete
    Async_issue_ -->|"yes (unhandled promise / race)"| Apply_Async_rules
    Async_issue_ -->|no| Error_handling_gap_
    Apply_Async_rules --> Code_complete
    Error_handling_gap_ -->|"yes (swallowed/untyped)"| Apply_Error_Handling_rules
    Error_handling_gap_ -->|no| Apply_Code_Quality_rules
    Apply_Error_Handling_rules --> Code_complete
    Apply_Code_Quality_rules --> Code_complete

Priority order: Type Safety > Async Correctness > Error Handling > Code Quality

Why These Rules Matter

Unhandled promise rejections: An async function called without await silently discarded its error in a payment handler. Node.js emitted UnhandledPromiseRejectionWarning — ignored in the logs — while orders were silently dropped. Fix: one missing await keyword.

any typing allowing runtime crashes: A third-party API response typed as any was accessed with .userId.toUpperCase(). The field was renamed to user_id in a patch release. TypeScript emitted no error; production crashed with Cannot read properties of undefined. Fix: type the response with unknown, validate the shape.

Missing null checks: A function returned undefined when no record was found, but the caller assumed it always returned a value. The non-null assertion ! had been added "because we know it's there". It wasn't, in one edge case. Fix: remove !, add a null check.

Silent promise failures in loops: for...of with await inside processed 500 items sequentially — 25 seconds instead of 1 second. The real bug was the developer's next change: wrapping the loop in Promise.all, which silently omitted the outer await, and the entire batch appeared to succeed while nothing ran. Fix: understand the async model before writing it.

These are real incidents. The rules exist because the pain is real.

Type Safety

TypeScript's value is catching errors at compile time. Rules that bypass the type system defeat the entire purpose of using TypeScript.

Never use any — use unknown with type guards:

// ❌ BAD: any silences the compiler entirely
function parseResponse(data: any): string {
    return data.user.name.toUpperCase();  // Crashes if shape changes
}

// ✅ GOOD: unknown forces you to verify the shape
function parseResponse(data: unknown): string {
    if (
        typeof data === "object" &&
        data !== null &&
        "user" in data &&
        typeof (data as { user: unknown }).user === "object"
    ) {
        const user = (data as { user: { name: string } }).user;
        return user.name.toUpperCase();
    }
    throw new Error("Unexpected response shape");
}

Avoid type assertions (as Type) — they suppress the compiler:

// ❌ BAD: Forces a type without verification
const user = response.data as User;
console.log(user.email.toLowerCase());  // Crashes if email is undefined

// ✅ GOOD: Parse and validate
function toUser(data: unknown): User {
    if (!isUser(data)) throw new Error("Invalid user shape");
    return data;
}

function isUser(data: unknown): data is User {
    return (
        typeof data === "object" &&
        data !== null &&
        typeof (data as Record<string, unknown>).email === "string"
    );
}

Avoid non-null assertions (!) — prefer optional chaining and nullish coalescing:

// ❌ BAD: Asserts non-null without evidence
const email = user!.profile!.email!.toLowerCase();

// ✅ GOOD: Safe navigation with fallback
const email = user?.profile?.email?.toLowerCase() ?? "unknown";

strict: true in tsconfig.json is non-negotiable. This enables strictNullChecks, noImplicitAny, strictFunctionTypes, and related checks. A codebase without strict: true is TypeScript in name only.

{
  "compilerOptions": {
    "strict": true,
    "noUncheckedIndexedAccess": true,
    "exactOptionalPropertyTypes": true
  }
}

Use discriminated unions for state modeling:

// ❌ BAD: Boolean flags create impossible states
interface FetchState {
    loading: boolean;
    error: boolean;
    data?: User;
}

// ✅ GOOD: Discriminated union — every state is explicit and valid
type FetchState =
    | { status: "idle" }
    | { status: "loading" }
    | { status: "error"; error: Error }
    | { status: "success"; data: User };

Prefer interface for object shapes, type for unions and intersections:

// Object shapes → interface (extensible, better error messages)
interface User {
    id: string;
    email: string;
}

// Unions and intersections → type
type ID = string | number;
type AdminUser = User & { role: "admin" };

Async Patterns

Async bugs are among the hardest to reproduce. They often only surface under load, on slow networks, or in specific timing windows.

Always await promises — unhandled rejections crash Node.js:

// ❌ BAD: Promise returned but not awaited — error silently discarded
async function saveOrder(order: Order): Promise<void> {
    db.insert(order);  // Missing await — failure is invisible
    sendConfirmationEmail(order.email);  // Also missing await
}

// ✅ GOOD: Explicit await — errors propagate correctly
async function saveOrder(order: Order): Promise<void> {
    await db.insert(order);
    await sendConfirmationEmail(order.email);
}

Never mix callbacks and promises without promisifying first:

// ❌ BAD: Error from callback not caught in promise chain
async function readFile(path: string): Promise<string> {
    return new Promise((resolve) => {
        fs.readFile(path, "utf-8", (err, data) => {
            if (err) throw err;  // Throws inside callback — unhandled
            resolve(data);
        });
    });
}

// ✅ GOOD: Rejection passed to the promise
async function readFile(path: string): Promise<string> {
    return new Promise((resolve, reject) => {
        fs.readFile(path, "utf-8", (err, data) => {
            if (err) reject(err);
            else resolve(data);
        });
    });
}

Use Promise.allSettled instead of Promise.all when partial failure is acceptable:

// ❌ BAD: One failure aborts all — other results lost
const results = await Promise.all(userIds.map(fetchUser));

// ✅ GOOD: Each settles independently
const results = await Promise.allSettled(userIds.map(fetchUser));
const users = results
    .filter((r): r is PromiseFulfilledResult<User> => r.status === "fulfilled")
    .map((r) => r.value);

Never await inside a loop unless sequential order is required:

// ❌ BAD: Sequential when parallel is possible — 500ms × N items
for (const id of userIds) {
    const user = await fetchUser(id);  // Waits for each before starting next
    results.push(user);
}

// ✅ GOOD: Parallel — all requests in flight simultaneously
const results = await Promise.all(userIds.map(fetchUser));

Error Handling

catch (e) gives unknown in strict mode — type-check before accessing:

// ❌ BAD: Accessing properties on unknown type
try {
    await processPayment(order);
} catch (e) {
    console.error(e.message);  // Error: 'e' is of type 'unknown'
}

// ✅ GOOD: Narrow the type before use
try {
    await processPayment(order);
} catch (e) {
    const message = e instanceof Error ? e.message : String(e);
    logger.error("Payment failed", { orderId: order.id, error: message });
    throw e;
}

Don't swallow errors silently — log or rethrow:

// ❌ BAD: Error swallowed — failure invisible to caller
try {
    await syncInventory();
} catch (_e) {
    // silently continue
}

// ✅ GOOD: Log and propagate — caller can handle or escalate
try {
    await syncInventory();
} catch (e) {
    logger.error("Inventory sync failed", { error: e });
    throw e;
}

Use discriminated union Result types for expected failures (not exceptions):

// For expected error paths (validation, not-found), prefer Result types over exceptions
type Result<T, E = Error> =
    | { ok: true; value: T }
    | { ok: false; error: E };

async function findUser(id: string): Promise<Result<User, "not-found" | "db-error">> {
    try {
        const user = await db.users.findById(id);
        if (!user) return { ok: false, error: "not-found" };
        return { ok: true, value: user };
    } catch {
        return { ok: false, error: "db-error" };
    }
}

// Callers must handle both paths — the type enforces it
const result = await findUser(id);
if (!result.ok) {
    if (result.error === "not-found") return Response.notFound();
    return Response.serverError();
}
return Response.ok(result.value);

Separate expected errors from unexpected errors. Network timeouts and validation failures are expected — use Result types. Programmer errors and invariant violations are unexpected — throw and let them crash loudly.

Testing

Preferred stack:

• Jest or Vitest — both support TypeScript natively with minimal config
• @testing-library/* — for UI component tests; test what the user sees, not implementation
• ts-expect or expectTypeOf (Vitest) — for type-level assertions
• msw (Mock Service Worker) — for HTTP mocking at the network level; prefer over manual fetch mocks
• testcontainers — for real database/service integration tests

Prefer real implementations over mocks for integration tests:

// ❌ BAD: Mock hides real integration surface
jest.mock("../db/userRepository");
const mockRepo = userRepository as jest.Mocked<typeof userRepository>;
mockRepo.findById.mockResolvedValue({ id: "1", email: "[e-mail]" });

// ✅ GOOD: In-memory implementation matches production contract
class InMemoryUserRepository implements UserRepository {
    private users = new Map<string, User>();
    async findById(id: string): Promise<User | null> {
        return this.users.get(id) ?? null;
    }
    async save(user: User): Promise<void> {
        this.users.set(user.id, user);
    }
}

Test the types too — use type-level assertions:

import { expectTypeOf } from "vitest";

test("parseUser returns User type", () => {
    const result = parseUser({ id: "1", email: "[e-mail]" });
    expectTypeOf(result).toEqualTypeOf<User>();
});

Don't test implementation details — test behavior:

// ❌ BAD: Tests internal structure — breaks on refactor
expect(service["_cache"].size).toBe(1);

// ✅ GOOD: Tests observable behavior
const result = await service.getUser("1");
expect(result).toEqual({ id: "1", email: "[e-mail]" });

Use beforeEach to reset state; never share mutable state between tests:

// ❌ BAD: Shared state causes test-order dependencies
const repo = new InMemoryUserRepository();

// ✅ GOOD: Fresh instance per test
let repo: InMemoryUserRepository;
beforeEach(() => {
    repo = new InMemoryUserRepository();
});

⛔ Bug Fix Workflow — Mandatory

See ~/.hortora/garden/approaches/testing.md — Bug Fix Workflow section for the mandatory 5-step process.

Code Quality

Mark parameters and variables readonly where possible:

// ❌ BAD: Mutable parameter — caller's array could be mutated
function processUsers(users: User[]): void {
    users.sort((a, b) => a.name.localeCompare(b.name));  // Mutates caller's array
}

// ✅ GOOD: readonly prevents accidental mutation
function processUsers(users: readonly User[]): void {
    const sorted = [...users].sort((a, b) => a.name.localeCompare(b.name));
    // ...
}

Prefer const over let, never var:

// ❌ BAD
var count = 0;
let message = "hello";  // message is never reassigned

// ✅ GOOD
const count = 0;
const message = "hello";

Use template literals over string concatenation:

// ❌ BAD
const url = "https://api.example.com/users/" + userId + "/orders/" + orderId;

// ✅ GOOD
const url = `https://api.example.com/users/${userId}/orders/${orderId}`;

Avoid deeply nested callbacks — use async/await:

// ❌ BAD: Callback pyramid
fetchUser(id, (user) => {
    fetchOrders(user.id, (orders) => {
        fetchInvoices(orders[0].id, (invoices) => {
            processAll(user, orders, invoices);
        });
    });
});

// ✅ GOOD: Linear async flow
const user = await fetchUser(id);
const orders = await fetchOrders(user.id);
const invoices = await fetchInvoices(orders[0].id);
processAll(user, orders, invoices);

Refactoring — IntelliJ First

Prerequisites: ide-tooling — invoke it for the full IntelliJ MCP tool guide. Always prefer IntelliJ MCPs over bash for any rename, move, find-references, or navigation. If no MCP is available for a semantic operation, inform the user — do not silently fall back.

Common Pitfalls — These Thoughts Mean STOP

If you catch yourself thinking any of these, STOP and apply the correct approach:

Rationalization	Problem	Impact	Fix
"TypeScript is just JavaScript, type errors are fine"	Type errors signal real bugs	Runtime crashes that strict mode prevents	Fix the type error; don't suppress it
"I'll add proper types later"	Technical debt never gets paid	any spreads — typed code calls untyped code	Type it now while context is fresh
"as any fixes it quickly"	Defeats the entire type system	Silent runtime failures TypeScript was meant to catch	Use unknown and a type guard
"Promise errors are handled somewhere up the chain"	They're not unless explicitly propagated	UnhandledPromiseRejection crashes Node.js	Explicitly await and catch
"await in a loop is fine for now"	Sequential when parallel is correct	50× performance regression	Use Promise.all
"I'll mock the database in tests"	Mock diverges from real behavior	Tests pass, production burns	Use in-memory implementation or real DB
"catch (e) { console.log(e) }"	Error swallowed after logging	Caller doesn't know it failed; state corrupted	Log AND rethrow
"Non-null assertion is safe here"	Crashes on edge cases	Cannot read properties of undefined in prod	Use optional chaining and null check
"I know this is a string" (using as string)	Runtime value may not be	Silent failure — wrong type passed downstream	Use a type guard to verify
"The types are too complex, I'll use any"	Complexity is a design smell	Cascading any propagates through the codebase	Simplify the design or use unknown
"This works at runtime, types are just noise"	Types are a correctness contract	The next person breaks it without warning	Fix the types to match the runtime
"@ts-ignore just for now"	Suppresses errors without fixing them	Real bug hidden until it surfaces in production	Fix the root cause instead

Prerequisites

Load ~/.hortora/garden/approaches/testing.md before proceeding. Apply all principles from that file (including the Playwright mandate for web UI).

Also apply all rules from ide-tooling: IntelliJ MCP tool guide — which tool to use for rename, move, find-references, navigation, diagnostics.

Skill Chaining

• Before committing: invoke code-review to catch type safety, async, and error handling issues
• For security-critical code: invoke security-audit when handling authentication, authorization, payment, or PII
• For dependency updates: invoke dependency-update when adding or upgrading packages
• For architectural decisions: suggest running adr to document significant design choices