fix(environment): measure the credential scan limit in UTF-8 bytes

A body counts against the 4 MiB limit by its UTF-8 length, not its UTF-16 code units, so multibyte text over the
limit is reported as unscanned instead of parsed.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

docs/specs/agent-environments/environment-build/spec.md

@@ -53,11 +53,11 @@ before building.
   parameters and no text, or the null text mitmproxy's SaveHar records, is
   scanned by those parameters, and a scenario transaction that cannot be
   attributed to one response still has its request side scanned. Text bodies
-  over the 4 MiB scan limit, including a response body decoded only as a prefix,
-  are not scanned; they are reported as a `credential-scan-incomplete` warning
-  with a count and the first 20 record and side pointers (`record 3 response`),
-  never contents, and never refuse a build. Binary bodies hold no JSON or form
-  text and are neither scanned nor reported.
+  over the 4 MiB scan limit, measured in UTF-8 bytes, including a response body
+  decoded only as a prefix, are not scanned; they are reported as a
+  `credential-scan-incomplete` warning with a count and the first 20 record and
+  side pointers (`record 3 response`), never contents, and never refuse a build.
+  Binary bodies hold no JSON or form text and are neither scanned nor reported.
 - The inspected HAR's Evidence and the projected manifest are admitted against
   the live heap headroom for parsing and sealing before either is serialized, as
   a capture Evidence file is before it is read; one that does not fit is refused

src/application/environment/environmentBuildReport.test.ts

@@ -287,6 +287,22 @@ describe("scanCredentialNames over structured and unattributed records", () => {
 });
 
 describe("scanCredentialNames coverage", () => {
+  it("measures the scan limit in UTF-8 bytes, not UTF-16 code units", () => {
+    // 1.5 million three-byte characters: under 4 Mi code units, over 4 MiB of UTF-8.
+    const wide = JSON.stringify({
+      token: "t-wide",
+      pad: "\u754c".repeat(1_500_000),
+    });
+    const multibyte = record(9, undefined, undefined);
+    (
+      multibyte.reported as { request: { postData?: unknown } }
+    ).request.postData = { mimeType: "application/json", text: wide };
+    expect(scanCredentialNames([multibyte])).toMatchObject({
+      count: 0,
+      skipped: { count: 1, pointers: ["record 9 request"] },
+    });
+  });
+
   it("reports text bodies too large to scan by record and side, never by content", () => {
     const large = JSON.stringify({
       token: "t-large",

src/application/environment/environmentBuildReport.ts

@@ -96,7 +96,13 @@ type ScannedBody =
 /** A recorded body as JSON or form fields, skipped when it is cut or over the limit, or undefined when it is neither. */
 const parsedBody = (text: unknown, cut = false): ScannedBody | undefined => {
   if (typeof text !== "string") return undefined;
-  if (cut || text.length > SCAN_BODY_LIMIT_BYTES) return { kind: "skipped" };
+  // UTF-8 is never shorter than UTF-16 code units, so the cheap length test only short-circuits.
+  if (
+    cut ||
+    text.length > SCAN_BODY_LIMIT_BYTES ||
+    Buffer.byteLength(text, "utf8") > SCAN_BODY_LIMIT_BYTES
+  )
+    return { kind: "skipped" };
   const start = text.trimStart()[0];
   if (start === "{" || start === "[") {
     try {