fix(run): add explicit UTF-8 encoding to prompt file operations (#604)

[sergio-sisternes-epam]

Summary\n\nFixes #604 — apm run start crashes with UnicodeDecodeError when .prompt.md contains non-ASCII characters on Windows systems with non-UTF-8 locale encoding (CP950, CP936, CP932).\n\n## Root Cause\n\nThree open() calls in script_runner.py did not pass an explicit encoding parameter:\n- PromptCompiler.compile() — reads the source .prompt.md\n- PromptCompiler.compile() — writes the compiled output file\n- ScriptRunner._execute_script() — reads the compiled file\n\nWithout encoding=\"utf-8\", Python uses the platform default (CP950 on affected systems), which cannot decode UTF-8 multi-byte sequences.\n\n## Fix\n\nAdded encoding=\"utf-8\" to all three open() calls. Full audit of the codebase confirmed no other prompt-related open() calls are affected.\n\n## Changes\n\n- src/apm_cli/core/script_runner.py — Add encoding=\"utf-8\" to 3 open() calls\n- tests/unit/test_script_runner.py — Add 2 tests for CJK and Cyrillic content in prompt compilation\n- CHANGELOG.md — Add fix entry\n\n## Testing\n\nAll 3792 tests pass. New tests verify:\n- CJK characters with frontmatter and parameter substitution\n- Cyrillic + emoji content without frontmatter\n

[Copilot]

Pull request overview

Fixes a Windows-specific UnicodeDecodeError in apm run start by forcing prompt compilation/read paths to use UTF-8, and adds regression tests + a changelog entry.

Changes:

• Add encoding="utf-8" to prompt read/write open() calls in the script runner prompt compilation flow.
• Add unit tests covering prompt compilation with non-ASCII content.
• Add an Unreleased changelog entry for the fix.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 4 comments.

File	Description
src/apm_cli/core/script_runner.py	Forces UTF-8 for reading source prompts, writing compiled output, and reading compiled prompts to avoid Windows locale decode errors.
tests/unit/test_script_runner.py	Adds regression tests for compiling prompts containing non-ASCII characters.
CHANGELOG.md	Records the fix under ## [Unreleased] -> ### Fixed.

[Copilot]

This test introduces non-ASCII characters directly in a Python source file (CJK and accented Latin in string literals). Repo encoding rules require all source files to remain printable ASCII; please represent these characters via Unicode escape sequences (e.g., \uXXXX/\UXXXXXXXX) while still writing UTF-8 bytes to the prompt file, so the test coverage remains but the source stays ASCII-only.

[Copilot]

This test includes Cyrillic text and an emoji in the Python source string literal. The repo encoding rules require Python source files to be printable ASCII only; please rewrite the literal using Unicode escape sequences (e.g., \u041f\u0440... and \U0001F680) to keep the source ASCII while still exercising UTF-8 file I/O.

[Copilot]

These tests manually manage a temp directory via tempfile.mkdtemp()/shutil.rmtree(). The rest of this file already uses tempfile.TemporaryDirectory() context managers; switching these new tests to TemporaryDirectory() would reduce cleanup risk (e.g., if rmtree fails on Windows due to open handles) and match existing conventions.

[Copilot]

Changelog entries are expected to end with the PR number (e.g., '(#648)') rather than the linked issue number. Please update this line to reference the PR ID, and optionally mention the issue in the text (e.g., 'Fixes #604') if desired.

[sergio-sisternes-epam]

Closing this PR — there are already two community contributions addressing #604:

• fix: add encoding="utf-8" to prompt file open() calls in script_runner (Windows CP950) #607 by @edenfunf (minimal targeted fix — same 3 open() calls)
• Standardize Unicode encodings to explicitly use UTF-8 #635 by @dvmukul (broader UTF-8 standardization across the codebase)

Deferring to the existing community PRs. Thank you to both contributors!