ask fails with JSON parsing error after successful authentication (v1.0.0.28144)

Description

workiq ask consistently fails with a JSON parsing error on Windows, including with a minimal prompt that does not reference any files or request work data.

Reproduced on 9 and 10 September 2026 by running workiq.exe directly, without the host application's launcher wrapper.

Environment

• Work IQ CLI: 1.0.0.28144+10c4074955aee0affce923a5fb04d7ed22c5a09e
• Windows x64
• PowerShell 7.6.6
• Default agent: bizchat-as-gpt-scenario

Minimal reproduction

From the directory containing workiq.exe, with an authenticated account:

.\workiq.exe ask --json --verbose -q "Reply with OK only. Do not search or retrieve any work data."
$LASTEXITCODE

Actual result

{"isError":true,"error":"\u0027{\u0027 is invalid after a single JSON value. Expected end of data. Path: $ | LineNumber: 0 | BytePositionInLine: 666.","agentId":"bizchat-as-gpt-scenario","agentName":"bizchat-as-gpt-scenario","agentProvider":"Microsoft"}

The process returns exit code 0 despite isError: true.

Without --json, the same parsing error appears. Verbose output also reports:

Warning: No conversation ID found in response

Expected result

The prompt returns an answer, or a diagnostic error that identifies the failing operation. A failed request should return a nonzero exit code.

Additional observations

• Authentication tracing reports successful silent token acquisition through WAM.
• agents list and agents get-card succeed.
• The failure reproduces without --file-urls.
• The installed executable's SHA-256 matches the published Windows x64 binary for package version 1.0.0.

I have not captured the underlying HTTP response, so I cannot determine whether this is a service response issue or a client parsing issue.

Could you advise how to collect a request/correlation ID or a redacted diagnostic trace for this failure? I can provide additional diagnostics through an appropriate private support channel if needed.

Start by reproducing the documented PowerShell command with workiq.exe ask --json --verbose, then compare it with the non-JSON invocation and the successful agents list or agents get-card commands. Trace the response handling around the ask command and capture the underlying response or correlation ID if available. Done means a minimal authenticated prompt returns an answer, or failures provide a useful diagnostic and nonzero exit code.