capskip 1.1.0 → 1.3.0

package/docs/API_REFERENCE.md

@@ -23,8 +23,14 @@ method returns a `Promise`.
 | reCAPTCHA v3 | `recaptcha(sitekey, url, { version: 'v3' })` | `userrecaptcha` + `version=v3` |
 | Cloudflare Turnstile | `turnstile()` | `turnstile` |
 | GeeTest v3 (slide) | `geetest()` | `geetest` |
+| ALTCHA (proof-of-work) | `altcha()` | `altcha` |
+| Capy Puzzle (slide) | `capy()` | `capy` |
+| CaptchaFox (widget) | `captchafox()` | `captchafox` |
+| Friendly Captcha (proof-of-work) | `friendlyCaptcha()` | `friendly_captcha` |
 
-**Proxy** is supported for reCAPTCHA, Turnstile, and GeeTest — not for image captcha.
+**Proxy** is supported for every method except image captcha. For ALTCHA the
+proxy is used only for the `challengeUrl` fetch; for Capy, only for the
+puzzle-image fetch, which is the only request a Capy solve makes.
 
 ---
 
@@ -267,6 +273,345 @@ rather than `defaultTimeout`.
 
 ---
 
+## 6. ALTCHA — `altcha(url, { ... })`
+
+ALTCHA is not a recognition captcha. There is no image, audio or text to read:
+the site issues a proof-of-work challenge and the client must brute-force a
+number that satisfies it. A solve is therefore deterministic and cheap —
+typically milliseconds.
+
+### POST `/in.php`
+
+| Parameter | Type | Required | Description |
+|---|---|---|---|
+| `key` | string | Yes | CapSkip API key |
+| `method` | string | Yes | `altcha` |
+| `pageurl` | string | Yes | Full URL of the page the challenge came from |
+| `challenge_url` | string | One of the two | Endpoint CapSkip fetches the challenge from |
+| `challenge_json` | string | One of the two | The challenge document itself, as a JSON string |
+| `json` | int | No | `0` plain text, `1` JSON |
+| `proxy` | string | No | Proxy address — used **only** for the `challenge_url` fetch |
+| `proxytype` | string | No | Proxy type |
+
+Sending both challenge parameters is allowed: the inline `challengeJson` wins,
+because fetching would only re-obtain what you already supplied.
+
+### Getting the challenge
+
+Open DevTools → Network on the target page and look for the request the
+`<altcha-widget>` makes for its challenge (often something like
+`/altcha/challenge`). The request URL is your `challengeUrl`; its JSON response
+is your `challengeJson`.
+
+The widget attribute that names that endpoint depends on the widget version, so
+read the page source rather than assuming:
+
+| Widget | Attribute |
+|---|---|
+| v1 / v2 | `challengeurl="…"`, with a separate `challengejson="…"` for an inline challenge |
+| v3+ | `challenge="…"` — the same attribute takes either a URL or the challenge data |
+
+> **Challenges expire, and the window is short** — some sites inside two minutes.
+> Once expired, the site refuses the solution with a bare "verification failed"
+> that looks exactly like a wrong answer. Fetch the challenge immediately before
+> solving and submit the token promptly; do not fetch a batch in advance, and do
+> not hold a token while a user fills in a form.
+>
+> CapSkip refuses an already-expired inline challenge immediately rather than
+> burning CPU on a token that cannot work. If you passed `challengeUrl` and the
+> challenge expired while the job queued, it fetches a fresh one automatically.
+
+### SDK usage
+
+```js
+// CapSkip fetches the challenge for you
+const result = await solver.altcha('https://example.com/signup', {
+  challengeUrl: 'https://example.com/captcha/api/altcha/challenge',
+});
+
+// …or hand it the document you already have. No network request at all.
+const inline = await solver.altcha('https://example.com/signup', {
+  challengeJson: {
+    algorithm: 'SHA-256', challenge: '…', salt: '…',
+    signature: '…', maxnumber: 1000000,
+  },
+});
+
+result.token;    // the base64 payload to post back
+result.number;   // the counter that solved it
+result.code;     // the same string as token
+```
+
+`challengeJson` accepts an object (serialized for you) or a JSON string.
+
+`number` is reported for both ALTCHA generations. Their tokens differ — a legacy
+payload carries it as a top-level `number`, while a proof-of-work v2 payload has
+none, its counter sitting at `solution.counter` — so it is read from the server's
+own `solution` object, which reports both the same way.
+
+Post the token back in the form field the widget uses, named `altcha`:
+
+```js
+await fetch(SIGNUP_URL, {
+  method: 'POST',
+  body: new URLSearchParams({
+    email: 'someone@example.com',
+    altcha: result.token,
+  }),
+});
+```
+
+Do not re-encode, trim or re-order the token: it is base64 of a JSON document
+whose fields are covered by the server's HMAC signature, so any modification
+invalidates it. Some integrations read the payload from a JSON body field
+instead — check what the page's own submit sends and mirror it.
+
+Unlike GeeTest and reCAPTCHA this is CPU proof-of-work rather than a browser
+solve, so it uses `defaultTimeout`, not `recaptchaTimeout`.
+
+### Algorithms
+
+CapSkip supports the legacy scheme (SHA-1/256/384/512) and PoW v2 with PBKDF2 or
+SHA. **Argon2id and scrypt are refused**, not attempted: a task using one returns
+`ERROR_CAPTCHA_UNSOLVABLE` and is never retried. ALTCHA itself recommends PBKDF2
+as the default, so this affects a minority of sites.
+
+All three widget types (`native`, `checkbox`, `switch`) work — the distinction is
+purely visual and never reaches CapSkip.
+
+---
+
+## 7. Capy Puzzle — `capy(sitekey, url, { ... })`
+
+Capy is a slide puzzle: a piece has to be dragged into the hole cut out of a
+photograph. CapSkip locates the hole and produces the drag path — one HTTP fetch
+plus pixel math, no browser.
+
+### POST `/in.php`
+
+| Parameter | Type | Required | Description |
+|---|---|---|---|
+| `key` | string | Yes | CapSkip API key |
+| `method` | string | Yes | `capy` |
+| `captchakey` | string | Yes | The site's public Capy key, conventionally prefixed `PUZZLE_`. The SDK sends its `sitekey` argument here. |
+| `pageurl` | string | Yes | Full URL of the page the captcha is on |
+| `api_server` | string | No | Root of the Capy API the key lives behind. Default `https://jp.api.capy.me` |
+| `version` | string | No | `puzzle` (the default). Only the puzzle family is solved |
+| `useragent` | string | No | User-Agent to send with the puzzle request |
+| `json` | int | No | `0` plain text, `1` JSON |
+| `proxy` | string | No | Proxy address |
+| `proxytype` | string | No | Proxy type |
+
+`sitekey` and `websiteKey` are accepted by the server as aliases for
+`captchakey`; the SDK sends the documented name.
+
+`version: 'avatar'` is refused by the SDK before the request is made. Avatar is a
+different challenge behind a different endpoint, and answering it with a puzzle
+answer would bill for a solve the target site rejects, indistinguishably from a
+broken solver.
+
+> `api.capy.me` no longer resolves, although several solver services still
+> document it. The live host is `jp.api.capy.me`; if a site's widget points
+> somewhere else, pass that as `api_server`.
+
+### Getting the captcha key
+
+It is in the page source as `capy_captchakey`, or in the widget script URL:
+`<script src="https://jp.api.capy.me/puzzle/get_js/?k=PUZZLE_XXXX">`.
+
+### GET `/res.php`
+
+The answer is **not a token**. It is an object, which the SDK expands into three
+result fields. See [Return value](#return-value).
+
+### SDK usage
+
+```js
+const result = await solver.capy(
+  'PUZZLE_Abc1dEFghIJKLM2no34P56q7rStu8v',
+  'https://example.com/login',
+);
+
+// With a non-default API server
+const result = await solver.capy(
+  'PUZZLE_Abc1dEFghIJKLM2no34P56q7rStu8v',
+  'https://example.com/login',
+  { api_server: 'https://jp.api.capy.me/' },
+);
+
+// capy_captchakey, capy_challengekey, capy_answer
+result.captchakey, result.challengekey, result.answer;
+```
+
+### Timing
+
+Capy grades the wall-clock gap between issuing the puzzle and verifying the
+answer, and refuses anything superhuman with `CAPTCHA verification failed` — the
+same message a wrong answer gets. CapSkip therefore holds every result until two
+seconds have elapsed since it drew the puzzle. This is invisible if you poll: the
+task simply takes about two seconds instead of a fifth of one. Nothing to
+configure.
+
+The challenge key is single-use and short-lived, so submit the three values
+promptly rather than caching them.
+
+---
+
+## 8. CaptchaFox — `captchafox(sitekey, url, { ... })`
+
+CaptchaFox scores the browser itself rather than asking the visitor to read
+anything; most visitors never see a puzzle. CapSkip drives the real widget in a
+real browser and returns the verification token.
+
+### POST `/in.php`
+
+| Parameter | Type | Required | Description |
+|---|---|---|---|
+| `key` | string | Yes | CapSkip API key |
+| `method` | string | Yes | `captchafox` |
+| `sitekey` | string | Yes | The public key the widget renders with, conventionally prefixed `sk_` |
+| `pageurl` | string | Yes | Full URL of the page the widget appears on |
+| `api_server` | string | No | Widget entry point. Default `https://cdn.captchafox.com/` |
+| `useragent` | string | No | Accepted for compatibility and not applied — CapSkip uses its own browser's identity |
+| `json` | int | No | `0` plain text, `1` JSON |
+| `proxy` | string | No | Proxy address |
+| `proxytype` | string | No | Proxy type |
+
+### Choosing the widget source
+
+| `api_server` | Token | Default |
+|---|---|---|
+| `https://cdn.captchafox.com/` | Plain | Yes |
+| `https://s.uicdn.com/mampkg/…` | `MAM_` prefixed | No |
+
+Read the value from the `<script>` tag that loads the widget. If you send the
+wrong one the solve still succeeds, but the token comes back in a format the
+target site will not accept — a silent verification failure rather than an error.
+
+### The page URL has to match the key
+
+CaptchaFox keys are registered against a list of allowed domains and the service
+checks the host before issuing anything. A correct key used on a page outside
+that list is refused permanently, not intermittently. CapSkip reports that case
+rather than retrying it, because retrying cannot help.
+
+### Challenge types
+
+| Challenge | Frequency | Supported |
+|---|---|---|
+| Invisible | Usually | Yes |
+| Slide | Sometimes | Yes |
+| Image select | Rarely | No |
+| Audio | Rarely | No |
+
+The two unsupported ones are uncommon and a retry usually draws a different
+challenge, so treat an unsolvable result as a signal to resubmit rather than a
+permanent failure of the key.
+
+### SDK usage
+
+```js
+const result = await solver.captchafox(
+  'sk_xtNxpk6fCdFbxh1_xJeGflSdCE9tn99G',
+  'https://example.com/signup',
+);
+
+// Post in the form field named cf-captcha-response
+result.token;
+
+// Submit under the UA that minted the token, not your own
+result.userAgent;
+```
+
+Uses `recaptchaTimeout` — it is a real browser session, and longer again when an
+interactive challenge is drawn.
+
+---
+
+## 9. Friendly Captcha — `friendlyCaptcha(sitekey, url, { ... })`
+
+Proof-of-work, with nothing shown on screen. The widget searches for values that
+hash below a difficulty the service sets per request.
+
+### POST `/in.php`
+
+| Parameter | Type | Required | Description |
+|---|---|---|---|
+| `key` | string | Yes | CapSkip API key |
+| `method` | string | Yes | `friendly_captcha` |
+| `sitekey` | string | Yes | The `data-sitekey` of the element carrying `class="frc-captcha"` |
+| `pageurl` | string | Yes | Full URL of the page the widget appears on |
+| `version` | string | No | `v1` (default) or `v2`; a bare `1` or `2` is accepted |
+| `module_script` | string | No | `src` of the widget script tag carrying `type="module"` |
+| `nomodule_script` | string | No | `src` of the widget script tag carrying `nomodule` |
+| `api_server` | string | No | Data residency endpoint: `global` (default), `eu`, or a full URL |
+| `useragent` | string | No | User-Agent to send with the request |
+| `json` | int | No | `0` plain text, `1` JSON |
+| `proxy` | string | No | Proxy address |
+| `proxytype` | string | No | Proxy type |
+
+### Version 1 and version 2
+
+Two entirely different protocols ship under this one name, and a sitekey does not
+tell you which one a site uses — they share a brand and a sitekey namespace and
+nothing else. Both are live. Solve the wrong one and you get a well-formed token
+the target site rejects, with no indication that the version was the problem.
+
+| `version` | Widget package | Script |
+|---|---|---|
+| `v1` | `friendly-challenge` | `widget.module.min.js` / `widget.min.js` |
+| `v2` | `@friendlycaptcha/sdk` | `site.min.js` |
+
+CapSkip decides in this order, stopping at the first answer: the `version`
+parameter; then the script URL from `moduleScript` or `nomoduleScript`, which is
+the most reliable signal because it is the build the site actually loads; then
+v1. The SDK refuses any other `version` value locally.
+
+### Submitting the token
+
+| `version` | Form field |
+|---|---|
+| `v1` | `frc-captcha-solution` |
+| `v2` | `frc-captcha-response` |
+
+The field names differ, which is what catches an integration moved from one to
+the other. A v1 token is four dot-separated parts and runs to a few hundred
+characters; a v2 token is a single opaque string beginning `AQQA.` and is roughly
+six kilobytes, so size whatever carries it accordingly.
+
+### SDK usage
+
+```js
+// Say which version
+const result = await solver.friendlyCaptcha(
+  'FCMGEMUD2M567T8G',
+  'https://example.com/signup',
+  { version: 'v2' },
+);
+
+// …or let the script URL decide it
+const result = await solver.friendlyCaptcha(
+  'FCMGEMUD2M567T8G',
+  'https://example.com/signup',
+  { moduleScript: 'https://cdn.example.com/@friendlycaptcha/sdk@0.1.6/site.min.js' },
+);
+
+// EU data-residency tenant
+const result = await solver.friendlyCaptcha(
+  'FCMGEMUD2M567T8G',
+  'https://example.com/signup',
+  { version: 'v2', api_server: 'eu' },
+);
+
+result.token;
+```
+
+Uses `recaptchaTimeout`: the service decides how much work a request is worth at
+the moment it is made, so solve time is not a constant, and v2 always solves in a
+browser.
+
+---
+
 ## Return value
 
 Every solve method resolves to:
@@ -279,6 +624,45 @@ Every solve method resolves to:
 }
 ```
 
+ALTCHA additionally exposes `token` (the same string as `code`, named for the
+form field it goes in) and `number`, the counter that solved it:
+
+```js
+{
+  captchaId: '12345',
+  code: 'eyJhbGdvcml0aG0iOiJTSEEtMjU2Iiwi…',
+  token: 'eyJhbGdvcml0aG0iOiJTSEEtMjU2Iiwi…',
+  number: 9661,
+}
+```
+
+CaptchaFox and Friendly Captcha expose `token`, the same string as `code`, named
+for the form field it goes in. CaptchaFox adds `userAgent` when the solve
+reported one — the UA the browser minted the token under, not one you sent:
+
+```js
+{
+  captchaId: '12345',
+  code: '177f50c25b845601e5c779cdb51b040d…',
+  token: '177f50c25b845601e5c779cdb51b040d…',
+  userAgent: 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) …',
+}
+```
+
+Capy returns an object rather than a token, expanded into the three fields the
+target form takes (`code` keeps the raw answer):
+
+```js
+{
+  captchaId: '12345',
+  code: { captchakey: 'PUZZLE_…', challengekey: '…', answer: '…', respKey: '' },
+  captchakey: 'PUZZLE_Abc1dEFghIJKLM2no34P56q7rStu8v',
+  challengekey: 'BalY2gJaI8uA2SGVOZhqBQ3V0CYSNNGP',
+  answer: '0xax8ex0xax84x0xkx7qx0x18x76x…',
+  respKey: '',
+}
+```
+
 GeeTest additionally expands its answer into `challenge`, `validate`, and
 `seccode` (`code` keeps the raw JSON string):
 
@@ -307,6 +691,12 @@ Convenience aliases mapped before sending to CapSkip:
 | `data_s` | `data-s` |
 | `apiServer` | `api_server` |
 | `api_subdomain` | `api_server` |
+| `challengeUrl` / `challengeURL` | `challenge_url` |
+| `challengeJson` / `challengeJSON` | `challenge_json` |
+| `userAgent` / `user_agent` | `useragent` |
+| `captchaKey` | `captchakey` |
+| `moduleScript` | `module_script` |
+| `nomoduleScript` / `noModuleScript` | `nomodule_script` |
 | `proxy` object | `proxy` + `proxytype` strings |
 
 ```js

package/examples/altcha.js

@@ -0,0 +1,103 @@
+'use strict';
+
+/**
+ * Solve an ALTCHA proof-of-work challenge.
+ *
+ * ALTCHA is not a recognition captcha -- there is nothing to read. The site
+ * issues a challenge and the browser must brute-force a number that satisfies
+ * it. CapSkip does that work for you, in milliseconds.
+ *
+ * You need the challenge, in one of two forms:
+ *
+ *   * `challengeUrl`  - the endpoint that serves it; CapSkip fetches it for you
+ *   * `challengeJson` - the challenge document itself, if you already have it
+ *
+ * To find them, open DevTools -> Network on the target page and look for the
+ * request the `<altcha-widget>` makes for its challenge (often something like
+ * `/altcha/challenge`). The request URL is your `challengeUrl`; its JSON
+ * response is your `challengeJson`.
+ *
+ * Note the widget attribute that names the endpoint changed between versions:
+ * v1/v2 use `challengeurl="..."`, while v3+ uses `challenge="..."` for both a
+ * URL and inline data. Read the page source rather than assuming.
+ *
+ * Challenges expire fast -- some sites inside two minutes -- so fetch one
+ * immediately before solving and post the token promptly. An expired challenge
+ * is rejected with a bare "verification failed" that looks exactly like a wrong
+ * answer.
+ *
+ * This example issues its own challenge the way a site's server would, so it
+ * runs as-is with no third-party dependency. Swap in your target's endpoint to
+ * use it for real.
+ */
+
+const crypto = require('crypto');
+
+const { CapSkip } = require('../src');
+
+const solver = new CapSkip({
+  apiKey: process.env.CAPSKIP_API_KEY || 'capskip',
+  host: process.env.CAPSKIP_HOST || '127.0.0.1',
+  port: Number(process.env.CAPSKIP_PORT || 8080),
+});
+
+const PAGE_URL = 'https://example.com/signup';
+
+/**
+ * Mint an ALTCHA challenge, exactly as a site's own server would.
+ *
+ * Replace this with a fetch of your target's challenge endpoint -- or skip it
+ * entirely and pass `challengeUrl` so CapSkip does the fetching.
+ */
+function issueChallenge(number = 54321) {
+  const salt = `${crypto.randomBytes(12).toString('hex')}?expires=${
+    Math.floor(Date.now() / 1000) + 600}`;
+  return {
+    algorithm: 'SHA-256',
+    challenge: crypto.createHash('sha256').update(`${salt}${number}`).digest('hex'),
+    salt,
+    signature: '0'.repeat(64),
+    maxnumber: 100000,
+  };
+}
+
+(async () => {
+  // --- Option A: you already have the challenge document ---------------------
+  // No network request at all: CapSkip solves it locally.
+  const result = await solver.altcha(PAGE_URL, { challengeJson: issueChallenge() });
+
+  console.log('Captcha ID:', result.captchaId);
+  console.log('Number:    ', result.number);
+  console.log('Token:     ', `${result.token.slice(0, 60)}...`);
+
+  // --- Option B: let CapSkip fetch the challenge ------------------------------
+  // Point it at the endpoint the widget calls. Add `proxy` if the endpoint
+  // should be fetched from a particular IP -- the proxy is used only for that
+  // fetch, never for the solve itself.
+  //
+  //   const result = await solver.altcha(PAGE_URL, {
+  //     challengeUrl: 'https://example.com/captcha/api/altcha/challenge',
+  //     proxy: { type: 'HTTP', uri: 'login:password@1.2.3.4:8080' },
+  //   });
+
+  // Post the token back in the form field the widget uses, named `altcha`:
+  //
+  //   await fetch(SIGNUP_URL, {
+  //     method: 'POST',
+  //     body: new URLSearchParams({
+  //       email: 'someone@example.com',
+  //       altcha: result.token,
+  //     }),
+  //   });
+  //
+  // Do not re-encode, trim or re-order it. The token is base64 of a JSON
+  // document whose fields are covered by the server's HMAC signature, so any
+  // modification invalidates it.
+
+  // `code` holds the same string as `token`, which is what you forward if you
+  // are porting code written against another solver's API.
+  console.assert(result.code === result.token);
+})().catch((err) => {
+  console.error(err);
+  process.exit(1);
+});

package/examples/captchafox.js

@@ -0,0 +1,102 @@
+'use strict';
+
+/**
+ * Solve a CaptchaFox challenge.
+ *
+ * CaptchaFox scores the browser itself rather than asking the visitor to read
+ * anything, so most solves draw no puzzle at all. CapSkip drives the real widget
+ * in a real browser and returns the verification token it would have produced.
+ *
+ * You need two values, both read off the target page:
+ *
+ *   * sitekey - the public key the widget renders with, conventionally prefixed
+ *     `sk_`. It is on the widget container as `data-sitekey`, or in the
+ *     `captchafox.render(...)` call. If the page builds the widget at runtime,
+ *     open DevTools -> Network and find the request to api.captchafox.com: the
+ *     key is the path segment after /captcha/.
+ *   * pageurl - the full URL of the page the widget appears on
+ *
+ * The page URL has to match: CaptchaFox keys are registered against a list of
+ * allowed domains and the service checks the host before it issues anything. A
+ * key that is correct but used on a page outside that list is refused
+ * permanently, not intermittently -- so if a sitekey fails immediately and
+ * consistently, check that you are sending the page the widget actually runs on
+ * rather than a search page, a redirect or a shortened link.
+ */
+
+const { CapSkip } = require('capskip');
+
+const solver = new CapSkip({
+  apiKey: process.env.CAPSKIP_API_KEY || 'capskip',
+  host: process.env.CAPSKIP_HOST || '127.0.0.1',
+  port: Number(process.env.CAPSKIP_PORT || 8080),
+});
+
+const SITEKEY = 'sk_xtNxpk6fCdFbxh1_xJeGflSdCE9tn99G';
+const PAGE_URL = 'https://example.com/signup';
+
+async function main() {
+  const result = await solver.captchafox(SITEKEY, PAGE_URL);
+
+  console.log('Captcha ID:', result.captchaId);
+  console.log('Token:     ', `${result.token.slice(0, 60)}...`);
+  console.log('User-Agent:', result.userAgent || '(none reported)');
+
+  // --- Choosing the widget source ------------------------------------------
+  // CaptchaFox publishes its widget from two places, and which one a site loads
+  // decides the shape of the token it expects back. Send api_server only when
+  // the target page does not use the default (https://cdn.captchafox.com/).
+  //
+  //   await solver.captchafox(SITEKEY, PAGE_URL, {
+  //     api_server: 'https://s.uicdn.com/mampkg/@mamdev/core.frontend.libs.captchafox/',
+  //   });
+  //
+  // That one returns a MAM_ prefixed token. Read the value from the <script>
+  // tag on the target page: if you send the wrong one the solve still succeeds,
+  // but the token comes back in a format the site will not accept, which looks
+  // like a silent verification failure rather than an error.
+  //
+  // --- With a per-request proxy --------------------------------------------
+  // CapSkip does not force a proxy, but CaptchaFox scores the network a widget
+  // runs on as well as the browser, so repeated solves from one address push
+  // that address toward the interactive challenges and then toward refusals.
+  //
+  //   await solver.captchafox(SITEKEY, PAGE_URL, {
+  //     proxy: { type: 'HTTP', uri: 'login:password@1.2.3.4:8080' },
+  //   });
+
+  // Post the token back in the form field the widget uses, named
+  // `cf-captcha-response`. Treat it as opaque and pass it through unchanged --
+  // it is verified server side against the session that produced it, so any
+  // edit invalidates it.
+  //
+  //   await fetch(SIGNUP_URL, {
+  //     method: 'POST',
+  //     headers: { 'User-Agent': result.userAgent },
+  //     body: new URLSearchParams({
+  //       email: 'someone@example.com',
+  //       'cf-captcha-response': result.token,
+  //     }),
+  //   });
+  //
+  // Some integrations read the token from a JSON body field instead, so check
+  // what the page's own submit sends and mirror it.
+  console.log('Form field:', { 'cf-captcha-response': `${result.token.slice(0, 40)}...` });
+
+  // Submit under the User-Agent the token was minted with, not your own.
+  // CapSkip solves in a real browser and uses that browser's identity, so the
+  // UA it reports is the one the token is bound to.
+  //
+  // Tokens are short lived: submit promptly rather than holding one while a
+  // user fills in a form, and solve again if the form is abandoned and resumed.
+
+  // Two of CaptchaFox's challenge types are not solvable -- image-select and
+  // the audio fallback, both rare. They are reported as unsolvable rather than
+  // waited out, so treat a failure as a signal to resubmit into a fresh
+  // challenge rather than as a permanent failure of the key.
+}
+
+main().catch((error) => {
+  console.error(`${error.name}: ${error.message}`);
+  process.exitCode = 1;
+});