owner can reset a member's platform password

The gap at the other end of create-user.ts: the owner could set a password once, at
creation, and never again. Losing it meant a hand-written UPDATE with an argon2
hash — the same "edit Postgres by hand" hole that creating accounts used to have.

POST /api/users/:id/password, owner-gated, with a button on the row.

GENERATED, not typed. The failure this exists for is "I created the account and
forgot to copy the password down", and an owner typing a replacement can lose it the
same way on the second go. Shown once in a dialog built to be copied — a dialog and
not a toast, because a toast that times out while somebody finds a pen loses the one
thing they came for.

The generator satisfies validatePassword BY CONSTRUCTION rather than by luck: one
character drawn from each of the four required classes, the rest from the union,
then Fisher-Yates shuffled so the first four positions are not always
lower/upper/digit/special. Rejection sampling throughout — `% n` on a byte biases
the early characters. Then it runs validatePassword on its own output, so if the
rules ever gain a requirement the alphabets do not cover it throws at the one call
site instead of minting passwords the login form rejects. Measured: 20,000
generations, all four classes present every time.

l, I, 1, O and 0 are absent from the alphabets. This gets read off a screen and
typed somewhere else.

Signs them out everywhere, as asked: passwordChangedAt = now, and userMiddleware
already refuses any token whose iat predates it. That overwrites the null
create-user leaves to mean "the owner chose this, not them" — checked, nothing reads
that column except the token check.

The Linux account is deliberately untouched, and the dialog says so. Members have no
Linux password and never had one: ensureOsUser runs useradd with no -p, so it is
created locked. Their terminal goes through setpriv, which does not authenticate;
their SSH is the key the owner pasted; `su - <member>` as root does not ask. And
machine-setup sets PasswordAuthentication no — verified on this host — so one could
not be used to log in even if it existed. Setting one would be a new way in, not a
repair.

The owner is excluded: they have change-password, which asks for the current one,
and resetting themselves here would end the session doing it.

Verified: transpiles, all lucide icons exist, 20k generator runs. tsgo next.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

src/servers/api/users/reset-user-password.ts

@@ -0,0 +1,121 @@
+import type { Handler } from 'hono';
+import { getUserById, updateUser, OWNER_USER_ID } from 'officerdb';
+import argon2 from 'argon2';
+import * as errors from '@@/custom-errors';
+import { validatePassword } from '../auth/validate-password';
+
+// The owner resetting somebody else's platform password.
+//
+// Named for what it is, and NOT to be confused with `api/auth/reset-password.ts`, which is the
+// token-from-an-email flow the person runs themselves. This one is the owner acting on another account,
+// and the only thing the two share is the column they write.
+//
+// ── Why the platform generates it rather than the owner typing one ──
+//
+// Because the failure this exists to fix is "I created the account and forgot to copy the password down".
+// An owner-typed reset can lose it exactly the same way on the second attempt. A generated one is shown
+// once, in a dialog built to be copied, and is never recoverable afterwards — which is the honest state,
+// since only the argon2 hash is stored.
+//
+// ── The Linux account is deliberately untouched ──
+//
+// A member's Linux account has no password and never had one: `ensureOsUser` runs `useradd` with no `-p`,
+// so it is created password-locked. Nothing needs one — their terminal runs through `setpriv`, which does
+// not authenticate; their SSH access is the key the owner pasted; `su - <member>` as root does not ask.
+// And machine-setup sets `PasswordAuthentication no`, so a Linux password could not be used to log in even
+// if it existed. Setting one here would be creating a new way in, not repairing a lost one.
+
+/**
+ * Characters the generated password is built from.
+ *
+ * Split by class because `validatePassword` requires ALL FOUR — twelve characters, a lowercase, an
+ * uppercase, a digit and a special. A uniform draw over the union satisfies that only with high
+ * probability, and "the reset button fails once a month" is a worse bug than it looks, because it happens
+ * to somebody who is already locked out.
+ *
+ * The specials are a subset of the class `validatePassword` accepts, chosen to survive being copied,
+ * pasted, quoted in a shell and read aloud: no quotes, no backslash, no backtick.
+ */
+const CLASSES = [
+  'abcdefghijkmnopqrstuvwxyz',
+  'ABCDEFGHJKLMNPQRSTUVWXYZ',
+  '23456789',
+  '!@#$%^&*()-_=+',
+] as const;
+
+const PASSWORD_LENGTH = 20;
+
+/** Uniform over `alphabet` by rejection sampling — `% n` on a byte biases the first characters. */
+function pick(alphabet: string): string {
+  const max = 256 - (256 % alphabet.length);
+  const byte = new Uint8Array(1);
+  for (;;) {
+    crypto.getRandomValues(byte);
+    if (byte[0]! < max) return alphabet[byte[0]! % alphabet.length]!;
+  }
+}
+
+/**
+ * A password that satisfies `validatePassword` by construction, not by luck.
+ *
+ * One character from each class first, the remainder from the union, then shuffled — otherwise the first
+ * four positions would always be lower, upper, digit, special, which is a pattern worth not having.
+ *
+ * Fisher-Yates with the same rejection sampling, for the same reason.
+ *
+ * `l`, `I`, `1`, `O` and `0` are absent from the alphabets above. This gets read off a screen and typed
+ * somewhere else often enough that the ambiguous pairs are worth losing.
+ */
+export function generatePassword(length = PASSWORD_LENGTH): string {
+  const union = CLASSES.join('');
+  const chars = CLASSES.map((cls) => pick(cls));
+  while (chars.length < length) chars.push(pick(union));
+
+  for (let i = chars.length - 1; i > 0; i--) {
+    const bound = i + 1;
+    const max = 256 - (256 % bound);
+    const byte = new Uint8Array(1);
+    let j: number;
+    do {
+      crypto.getRandomValues(byte);
+    } while (byte[0]! >= max);
+    j = byte[0]! % bound;
+    [chars[i], chars[j]] = [chars[j]!, chars[i]!];
+  }
+
+  const password = chars.join('');
+  // Generated then CHECKED against the same validator every other path uses. If the rules ever gain a
+  // requirement the alphabets above do not cover, this throws here — at the one call site, loudly — rather
+  // than minting passwords the login form would go on to reject.
+  validatePassword(password);
+  return password;
+}
+
+export const resetUserPasswordHandler: Handler = async function (ctx) {
+  const id = Number(ctx.req.param('id'));
+  if (!Number.isInteger(id) || id < 1) throw errors.BAD_REQUEST('Invalid user id');
+
+  // The owner has `PUT /auth/change-password`, which asks for the current one. Resetting themselves from
+  // here would sign them out of the session they are doing it from, for no gain.
+  if (id === OWNER_USER_ID) {
+    throw errors.BAD_REQUEST('Change your own password from Settings, which asks for the current one.');
+  }
+
+  const existing = await getUserById(id);
+  if (!existing) throw errors.NOT_FOUND('User not found');
+
+  const password = generatePassword();
+
+  // `passwordChangedAt: new Date()` is what signs them out everywhere. `userMiddleware` refuses any token
+  // whose `iat` predates it, so every session they have — browser, phone, an agent holding a token —
+  // stops at the next request. That is the point: a reset happens because the old password was lost or
+  // leaked, and leaving live sessions behind would defeat both reasons.
+  //
+  // It also overwrites the null that `create-user.ts` leaves to mean "the owner chose this, not them".
+  // Nothing reads that; the only consumer of this column is the token check above.
+  await updateUser(id, { password: await argon2.hash(password), passwordChangedAt: new Date() });
+
+  // Returned exactly once. Only the hash is stored, so there is no second chance to fetch it and the UI
+  // says so — which is the whole reason it is generated rather than typed.
+  return ctx.json({ email: existing.email, password });
+};