Fix/windows utf8 io

[Le-Anh-Duy]

@QingtaoLi1 Thanks for checking! I dug into this and can confirm there's a real, reproducible Windows bug — just not exactly where I first assumed.

Confirmed: cmind script update_graphs.py status (used by the SessionStart/post-commit hooks) crashes with UnicodeEncodeError on Windows whenever an RPG already exists:

UnicodeEncodeError: 'charmap' codec can't encode character '→' in position ...

Root cause: on Windows, a script's stdout is almost never a real console (it's piped by cmind script, captured by a hook, etc.), so CPython falls back to locale.getpreferredencoding() for stdio instead of UTF-8 — a legacy code page (cp1252/cp936/...) on most Windows installs. _format_status_for_agent() unconditionally prints an MCP-tools guidance line containing "→" (functional areas → groups → features) whenever the RPG is readable, so this fires on the ordinary "status looks fine" path, not just some rare edge case.

I reproduced it end-to-end on a fresh 1-file sample repo: cmind script update_graphs.py status crashes on the unfixed build and succeeds on the fixed one, no LLM calls needed to trigger it.

Not confirmed: whether this exact code path is what broke /cmind.encode itself for you — run_encode.py/check_encode.py's own print(json.dumps(...)) calls are ASCII-safe by default (ensure_ascii=True), so they shouldn't hit this. If some other print()/log call deeper in the encode pipeline emitted a non-ASCII character (LLMs often produce "invisible" typographic characters — smart quotes, em-dashes, arrows, bullets — even for plain English content), it would crash the same way. Could you share the exact traceback/error text you saw from /cmind.encode? That'll tell us for sure whether it's the same bug or a second one.

Pushed a fix to fix/windows-utf8-io: forces UTF-8 stdio at the single import choke point every bundled script shares (common/paths.py), plus PYTHONIOENCODING/errors="replace" as defense in depth around the cmind script subprocess launch and the LLM CLI subprocess. Added a regression test and a from-scratch, LLM-free repro under tests/repro_windows_utf8/ if you want to verify on your machine.

[Copilot]

Pull request overview

This PR addresses a reproducible Windows UnicodeEncodeError caused by non-UTF-8 stdio when bundled scripts run with piped/captured output (common for hooks and cmind script). The fix forces UTF-8 stdio consistently across script entrypoints and subprocesses, and adds regression coverage plus a fast local repro.

Changes:

• Force UTF-8 stdio (with errors="replace") for bundled scripts and key subprocess launches to prevent Windows code-page encoding crashes.
• Add a regression test covering non-ASCII status output and a minimal, LLM-free Windows repro harness.
• Refactor process-tree termination logic to be cross-platform (POSIX process groups vs. Windows taskkill /T).

Reviewed changes

Copilot reviewed 8 out of 9 changed files in this pull request and generated 3 comments.

Show a summary per file

File	Description
CoderMind/tests/test_hooks_install.py	Adds regression test to ensure non-ASCII status output doesn’t crash under captured stdout.
CoderMind/tests/repro_windows_utf8/repro_runner.py	Adds a fast repro driver that mimics CoderMind’s subprocess capture pattern.
CoderMind/tests/repro_windows_utf8/repro_child.py	Adds a minimal child script that prints a Unicode arrow to reproduce the failure mode.
CoderMind/tests/repro_windows_utf8/README.md	Documents the Windows failure mode, evidence, and repro steps.
CoderMind/src/cmind_cli/init.py	Forces UTF-8 environment/decoding for script and encoder subprocesses.
CoderMind/scripts/common/paths.py	Reconfigures sys.stdout/sys.stderr to UTF-8 at a shared import “choke point”.
CoderMind/scripts/common/llm_client.py	Adds cross-platform process-tree killing and Windows-safe subprocess configuration for the AI CLI.
CoderMind/scripts/code_gen/test_runner.py	Switches pytest subprocess management to use the shared cross-platform kill logic.
.gitignore	Updates ignore rules for .cmind and related generated directories/files.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[QingtaoLi1]

@Le-Anh-Duy Thank you for your further efforts! I tried this new PR, and find another small patch that can make it work on my Windows device. The patch is as follows:

diff --git a/CoderMind/scripts/common/llm_client.py b/CoderMind/scripts/common/llm_client.py
index 6bfbad9..84387f3 100644
--- a/CoderMind/scripts/common/llm_client.py
+++ b/CoderMind/scripts/common/llm_client.py
@@ -168,8 +168,21 @@ def detect_agent_type(cmd: Optional[str] = None) -> str:
     if not cmd or cmd == _PLACEHOLDER_LITERAL:
         return "unknown"

-    first_token = cmd.strip().split()[0]
-    return _CLI_TO_AGENT.get(first_token, "unknown")
+    try:
+        first_token = shlex.split(cmd, posix=False)[0]
+    except (IndexError, ValueError):
+        parts = cmd.strip().split(maxsplit=1)
+        if not parts:
+            return "unknown"
+        first_token = parts[0]
+
+    executable_name = first_token.strip("\"'").replace("\\", "/").rsplit("/", 1)[-1].lower()
+    for suffix in (".cmd", ".exe", ".bat", ".ps1"):
+        if executable_name.endswith(suffix):
+            executable_name = executable_name[: -len(suffix)]
+            break
+
+    return _CLI_TO_AGENT.get(executable_name, "unknown")