[tangzhiqiang]

What changed

samples/src/main/resources/log4j2.properties — the console appender's layout now pins its charset
explicitly, matching the rolling file appender that already did:

 appender.console.layout.type=PatternLayout
 appender.console.layout.pattern=${LOG_PATTERN}
+appender.console.layout.charset=UTF-8
 # Rolling file appender

Why

appender.file.layout.charset was already UTF-8, but the console appender had no charset set, so
it fell back to the JVM default — on Windows that is the legacy code page (GBK on Chinese
systems). Non-ASCII log text therefore rendered as mojibake:

• task prompts and agent responses
• A2A-T protocol content — the SDK logs full protocol bodies at DEBUG through the PROTOCOL logger

The same text was correct in logs/spn-demo.log but garbled on stdout, and therefore also garbled
in any log collector that captures stdout instead of the file. Pinning the console charset makes the
two appenders consistent.

Breaking changes

None. No API, configuration surface, or module behaviour changes. The only observable difference is
that console log output is now UTF-8 instead of the platform default, which is the intent.

Test results

mvn -B -pl samples -am verify

• workflow-engine, spring-boot-starter — build green
• samples — 97 tests, with the same 2 failures / 4 errors in SpringSpnDemoE2ETest that
  reproduce on the pristine baseline without this commit. The failure is
  PortInUseException on an ephemeral port the E2E demo binds (62478 on one run, 61222 on
  another — i.e. the port is picked dynamically and the local Windows box refuses it, most likely a
  reserved port range). Identical counts on the baseline, so it is not a regression from this
  change. CI on ubuntu-latest is the authoritative gate for this module.

Docs

No documentation changes needed — this is a logging resource fix inside the samples module and
does not affect any documented API or configuration surface.