CVE-2026-113667

**Summary** Form and login CAPTCHA verification interpolate the user-supplied widget token unencoded into the server-side verification URL/body, allowing a crafted token to inject query parameters into the reCAPTCHA/hCaptcha verify request. **Details** - `packages/form/lib/recaptcha.js:39` — `recaptchaUri = 'https://www.google.com/recaptcha/api/siteverify?secret=${recaptchaSecret}&response=${input.recaptcha}'` - `packages/login-recaptcha/index.js:44` — same pattern for login - `packages/login-hcaptcha/index.js:47` — `body: 'response=${token}&secret=${secret}'` **PoC** ``` input.recaptcha = 'attacker_token&secret=INJECTED_SECRET&something=1' → resulting verify URL: https://www.google.com/recaptcha/api/siteverify?secret=<real>&response=attacker_token&secret=INJECTED_SECRET&something=1 # (replicated exactly from recaptcha.js:39) ``` **Impact** Low: a crafted token smuggles extra `key=value` pairs (`secret`, extra `response`) into the outbound verify call, forcing verification failures (DoS of the captcha gate and of captcha-gated login/form flows) and enabling automated bypass probing; no secret exfiltration. **Remediation** — Build the verify request with `URL` + `searchParams` (or `encodeURIComponent` on both `secret` and `response`), and validate the token against `^[A-Za-z0-9._-]{1,2048}$` before use.