bug: UnicodeDecodeError reading .prompt.md on Windows CP950 — open() missing encoding parameter

[edenfunf]

Description

apm run start crashes with a UnicodeDecodeError when the .prompt.md file contains any character outside ASCII on a Windows system with a non-UTF-8 locale encoding (CP950, CP936, CP932).

Steps to Reproduce

1. Windows machine with a CJK locale (console code page CP950)
2. apm init my-project && cd my-project
3. Edit start.prompt.md to include any CJK character (e.g. # 開始)
4. apm run start

Actual Behaviour

[x] Script execution error: 'cp950' codec can't decode byte 0x8b in position 12: illegal multibyte sequence

Expected Behaviour

The prompt file is read as UTF-8 and the command executes successfully.

Root Cause

Three open() calls in script_runner.py do not pass an explicit encoding parameter:

• PromptCompiler.compile() — reads the source .prompt.md
• PromptCompiler.compile() — writes the compiled output file
• ScriptRunner._execute_script() — reads the compiled file

Without encoding="utf-8", Python uses the platform default (CP950 on affected systems), which cannot decode UTF-8 multi-byte sequences.

Environment

• OS: Windows 11, console code page CP950
• apm version: v0.8.11