fix(cli): read files saved by Windows PowerShell's redirect

dist/cli/index.js

@@ -90,6 +90,7 @@ const run_1 = require("../sandbox/run");
 const guardrails_1 = require("../lib/guardrails");
 const api_url_1 = require("../lib/api-url");
 const install_key_1 = require("../lib/install-key");
+const decode_text_1 = require("../lib/decode-text");
 const owner_receipt_1 = require("../lib/owner-receipt");
 /** Exit codes — a small, stable contract so CI scripts can branch on them. */
 exports.EXIT = {
@@ -1373,7 +1374,7 @@ async function run(args, client, io = realIO) {
             const file = args.operand;
             let raw;
             try {
-                raw = (0, node_fs_1.readFileSync)(file, 'utf8');
+                raw = (0, decode_text_1.decodeText)((0, node_fs_1.readFileSync)(file));
             }
             catch (e) {
                 io.err(`verify-receipt: cannot read ${file}: ${e.message}`);
@@ -1399,7 +1400,7 @@ async function run(args, client, io = realIO) {
             let envelope = obj && typeof obj === 'object' ? obj['six_claims'] : undefined;
             if (args.envelope) {
                 try {
-                    envelope = JSON.parse((0, node_fs_1.readFileSync)(args.envelope, 'utf8'));
+                    envelope = JSON.parse((0, decode_text_1.decodeText)((0, node_fs_1.readFileSync)(args.envelope)));
                 }
                 catch (e) {
                     io.err(`verify-receipt: cannot read --envelope ${args.envelope}: ${e.message}`);
@@ -1426,7 +1427,7 @@ async function run(args, client, io = realIO) {
             const file = args.operand;
             let raw;
             try {
-                raw = (0, node_fs_1.readFileSync)(file === '-' ? 0 : file, 'utf8'); // fd 0 = stdin
+                raw = (0, decode_text_1.decodeText)((0, node_fs_1.readFileSync)(file === '-' ? 0 : file)); // fd 0 = stdin
             }
             catch (e) {
                 io.err(`sign-receipt: cannot read ${file === '-' ? 'stdin' : file}: ${e.message}`);
@@ -1683,7 +1684,7 @@ async function run(args, client, io = realIO) {
                         return exports.EXIT.USAGE;
                     }
                     try {
-                        evidence = JSON.parse(fs.readFileSync(args.evidence, 'utf8'));
+                        evidence = JSON.parse((0, decode_text_1.decodeText)(fs.readFileSync(args.evidence)));
                     }
                     catch {
                         io.err(`report: evidence file is not valid JSON: ${args.evidence}`);
@@ -1731,11 +1732,9 @@ function readClaimsFile(path, fs = require('node:fs')) {
         throw new Error(`--claims-file: no such file: ${path}`);
     if (st.size > action_evidence_1.MAX_CLAIMS_FILE_BYTES)
         throw new Error(`--claims-file is ${st.size} bytes; the limit is ${action_evidence_1.MAX_CLAIMS_FILE_BYTES}`);
-    const text = String(fs.readFileSync(path, 'utf8'));
-    // One leading byte-order mark is a file-encoding marker, not text: Windows PowerShell 5.1 writes it
-    // for `Out-File` and `Set-Content -Encoding utf8`, and kept it would make line 1 non-ASCII and
-    // block every verdict (XC1 re-check 5). Only that one character is dropped; any other stays.
-    return text.charCodeAt(0) === 0xfeff ? text.slice(1) : text;
+    // Windows PowerShell 5.1 writes a byte-order mark for `Out-File` and `Set-Content -Encoding utf8`, and UTF-16 for
+    // `>`. A mark kept in the text would make line 1 non-ASCII and block every verdict (XC1 re-check 5).
+    return (0, decode_text_1.decodeText)(fs.readFileSync(path));
 }
 /**
  * `trustshell check <runUrl> --claims "<text>"` (or `--claims-file <path>`). GitHub only, exactly like the URL form: the

dist/lib/decode-text.d.ts

@@ -0,0 +1,6 @@
+/**
+ * Text a person saved on any platform. Windows PowerShell 5.1's `>` writes UTF-16 with a byte-order mark, so
+ * `trustshell sign-receipt body.json > r.json`, the documented command, produced a file `verify-receipt` called
+ * "not valid JSON". A byte-order mark says how the bytes are encoded and is never part of the text.
+ */
+export declare function decodeText(raw: Buffer | string): string;

dist/lib/decode-text.js

@@ -0,0 +1,18 @@
+"use strict";
+Object.defineProperty(exports, "__esModule", { value: true });
+exports.decodeText = decodeText;
+/**
+ * Text a person saved on any platform. Windows PowerShell 5.1's `>` writes UTF-16 with a byte-order mark, so
+ * `trustshell sign-receipt body.json > r.json`, the documented command, produced a file `verify-receipt` called
+ * "not valid JSON". A byte-order mark says how the bytes are encoded and is never part of the text.
+ */
+function decodeText(raw) {
+    if (typeof raw === 'string')
+        return raw.charCodeAt(0) === 0xfeff ? raw.slice(1) : raw;
+    if (raw[0] === 0xff && raw[1] === 0xfe)
+        return raw.subarray(2).toString('utf16le');
+    if (raw[0] === 0xfe && raw[1] === 0xff)
+        return Buffer.from(raw.subarray(2)).swap16().toString('utf16le');
+    const text = raw.toString('utf8');
+    return text.charCodeAt(0) === 0xfeff ? text.slice(1) : text;
+}

src/cli/index.ts

@@ -78,6 +78,7 @@ import { sandboxCli } from '../sandbox/run';
 import { GUARD_RECORD_SCHEMA, trustshellHome } from '../lib/guardrails';
 import { ApiUrlError } from '../lib/api-url';
 import { signingKey } from '../lib/install-key';
+import { decodeText } from '../lib/decode-text';
 import { signOwnerReceipt, installKeyRootHex, type SealableReceipt } from '../lib/owner-receipt';
 
 /** Exit codes — a small, stable contract so CI scripts can branch on them. */
@@ -1454,7 +1455,7 @@ export async function run(
       const file = args.operand as string;
       let raw: string;
       try {
-        raw = readTextFile(file, 'utf8');
+        raw = decodeText(readTextFile(file));
       } catch (e) {
         io.err(`verify-receipt: cannot read ${file}: ${(e as Error).message}`);
         return EXIT.USAGE;
@@ -1477,7 +1478,7 @@ export async function run(
       let envelope: unknown = obj && typeof obj === 'object' ? obj['six_claims'] : undefined;
       if (args.envelope) {
         try {
-          envelope = JSON.parse(readTextFile(args.envelope, 'utf8'));
+          envelope = JSON.parse(decodeText(readTextFile(args.envelope)));
         } catch (e) {
           io.err(`verify-receipt: cannot read --envelope ${args.envelope}: ${(e as Error).message}`);
           return EXIT.USAGE;
@@ -1502,7 +1503,7 @@ export async function run(
       const file = args.operand as string;
       let raw: string;
       try {
-        raw = readTextFile(file === '-' ? 0 : file, 'utf8'); // fd 0 = stdin
+        raw = decodeText(readTextFile(file === '-' ? 0 : file)); // fd 0 = stdin
       } catch (e) {
         io.err(`sign-receipt: cannot read ${file === '-' ? 'stdin' : file}: ${(e as Error).message}`);
         return EXIT.USAGE;
@@ -1746,7 +1747,7 @@ export async function run(
             return EXIT.USAGE;
           }
           try {
-            evidence = JSON.parse(fs.readFileSync(args.evidence, 'utf8')) as EvidenceDoc;
+            evidence = JSON.parse(decodeText(fs.readFileSync(args.evidence))) as EvidenceDoc;
           } catch {
             io.err(`report: evidence file is not valid JSON: ${args.evidence}`);
             return EXIT.USAGE;
@@ -1796,11 +1797,9 @@ export function readClaimsFile(
   }
   if (!st?.isFile()) throw new Error(`--claims-file: no such file: ${path}`);
   if (st.size > MAX_CLAIMS_FILE_BYTES) throw new Error(`--claims-file is ${st.size} bytes; the limit is ${MAX_CLAIMS_FILE_BYTES}`);
-  const text = String(fs.readFileSync(path, 'utf8'));
-  // One leading byte-order mark is a file-encoding marker, not text: Windows PowerShell 5.1 writes it
-  // for `Out-File` and `Set-Content -Encoding utf8`, and kept it would make line 1 non-ASCII and
-  // block every verdict (XC1 re-check 5). Only that one character is dropped; any other stays.
-  return text.charCodeAt(0) === 0xfeff ? text.slice(1) : text;
+  // Windows PowerShell 5.1 writes a byte-order mark for `Out-File` and `Set-Content -Encoding utf8`, and UTF-16 for
+  // `>`. A mark kept in the text would make line 1 non-ASCII and block every verdict (XC1 re-check 5).
+  return decodeText(fs.readFileSync(path));
 }
 
 /**

src/lib/decode-text.ts

@@ -0,0 +1,12 @@
+/**
+ * Text a person saved on any platform. Windows PowerShell 5.1's `>` writes UTF-16 with a byte-order mark, so
+ * `trustshell sign-receipt body.json > r.json`, the documented command, produced a file `verify-receipt` called
+ * "not valid JSON". A byte-order mark says how the bytes are encoded and is never part of the text.
+ */
+export function decodeText(raw: Buffer | string): string {
+  if (typeof raw === 'string') return raw.charCodeAt(0) === 0xfeff ? raw.slice(1) : raw;
+  if (raw[0] === 0xff && raw[1] === 0xfe) return raw.subarray(2).toString('utf16le');
+  if (raw[0] === 0xfe && raw[1] === 0xff) return Buffer.from(raw.subarray(2)).swap16().toString('utf16le');
+  const text = raw.toString('utf8');
+  return text.charCodeAt(0) === 0xfeff ? text.slice(1) : text;
+}

tests/decode-text.test.ts

@@ -0,0 +1,18 @@
+import { decodeText } from '../src/lib/decode-text';
+
+describe('decodeText', () => {
+  const text = '{"a":"é"}\n';
+  it('reads what Windows PowerShell 5.1 > writes', () => {
+    expect(decodeText(Buffer.concat([Buffer.from([0xff, 0xfe]), Buffer.from(text, 'utf16le')]))).toBe(text);
+  });
+  it('reads UTF-16 big-endian', () => {
+    expect(decodeText(Buffer.concat([Buffer.from([0xfe, 0xff]), Buffer.from(text, 'utf16le').swap16()]))).toBe(text);
+  });
+  it('drops one UTF-8 byte-order mark and keeps the rest', () => {
+    expect(decodeText(Buffer.from(`${text}`, 'utf8'))).toBe(text);
+    expect(decodeText(`${text}`)).toBe(`${text}`);
+  });
+  it('leaves plain UTF-8 alone', () => {
+    expect(decodeText(Buffer.from(text, 'utf8'))).toBe(text);
+  });
+});