encoding-guard

name	encoding-guard
description	Prevent garbled text (乱码) when editing code and comments. Use this skill whenever reading, writing, or modifying files to ensure proper UTF-8 encoding handling. Activates on any file edit operation, especially when working with Chinese characters, emoji, or non-ASCII content.

Encoding Guard

Ensures all file operations use proper UTF-8 encoding to prevent garbled text (乱码).

Core Principles

1. Always use UTF-8 — Read and write files with explicit UTF-8 encoding
2. Validate before write — Check that content is valid UTF-8 before saving
3. Preserve BOM awareness — Detect and handle BOM (Byte Order Mark) when present
4. Handle mixed encoding — Convert legacy encodings (GBK, GB2312, Big5) to UTF-8

When Reading Files

# Always specify encoding when reading
cat -n file.tsx  # Use Read tool which handles encoding

# If you suspect encoding issues, check with:
file -I file.tsx  # Shows charset
hexdump -C file.tsx | head -5  # Inspect raw bytes

When Writing Files

# Ensure locale supports UTF-8
export LANG=en_US.UTF-8
export LC_ALL=en_US.UTF-8

# When using Edit tool, the content is already validated
# When using Write tool, ensure content is proper UTF-8

Detection Patterns

Watch for these signs of encoding problems:

1. Mojibake characters: é, è, ü (UTF-8 read as Latin-1)
2. Chinese garbled: 锟斤拷, 烫烫烫, 屯屯屯 (GBK/UTF-8 mismatch)
3. Replacement character: � (U+FFFD) — missing or invalid bytes
4. Question marks: ??? where Chinese should be

Fix Strategies

Strategy 1: Re-read with correct encoding

If a file shows garbled text:

# Check current encoding
file -I problematic.tsx

# If it's GBK, convert to UTF-8
iconv -f GBK -t UTF-8 problematic.tsx > fixed.tsx
mv fixed.tsx problematic.tsx

Strategy 2: Use Read tool with encoding awareness

The Read tool automatically handles encoding. If you see garbled output:

1. Stop and check the file's actual encoding
2. Convert if needed before editing
3. Then proceed with normal Read/Edit operations

Strategy 3: Validate after write

After any Edit or Write operation:

# Quick validation
python3 -c "
with open('file.tsx', 'r', encoding='utf-8') as f:
    content = f.read()
    print(f'Valid UTF-8: {len(content)} chars')
"

Common Scenarios

Scenario 1: Editing TypeScript/JSX with Chinese comments

// ✅ Correct: UTF-8 encoded comment
// 这是一个中文注释

// ❌ Wrong: Encoding mismatch causes garbled
// 杩欐槸涓€涓�腑鏂囨敞閲�

Action: Always use UTF-8 locale. Set in shell:

export LANG=en_US.UTF-8

Scenario 2: Mixed encoding in project

If some files are GBK and others UTF-8:

# Find all non-UTF-8 files
find . -name "*.tsx" -exec file -I {} \; | grep -v utf-8

# Batch convert
for f in $(find . -name "*.tsx" -exec file -I {} \; | grep gb2312 | cut -d: -f1); do
    iconv -f GBK -t UTF-8 "$f" > "$f.utf8"
    mv "$f.utf8" "$f"
done

Scenario 3: Git shows encoding warnings

# Configure git to handle encoding
git config core.quotepath false  # Show Chinese filenames correctly
git config i18n.commitencoding utf-8
git config i18n.logoutputencoding utf-8

Prevention Checklist

Before any file operation:

• Locale is set to UTF-8 (echo $LANG should show *.UTF-8)
• File encoding is verified (file -I filename)
• Content being written is valid UTF-8
• Editor/IDE is configured for UTF-8

Quick Diagnostic

Run this to check your environment:

echo "Locale: $LANG"
echo "LC_ALL: $LC_ALL"
python3 -c "import sys; print(f'Default encoding: {sys.getdefaultencoding()}')"
python3 -c "import locale; print(f'Preferred encoding: {locale.getpreferredencoding()}')"

Expected output:

Locale: en_US.UTF-8
LC_ALL: en_US.UTF-8
Default encoding: utf-8
Preferred encoding: UTF-8

Integration with Tools

Read Tool

• Automatically detects and handles encoding
• If garbled, the file may need conversion first

Edit Tool

• Content parameter must be valid UTF-8
• Tool validates encoding before applying changes

Write Tool

• Content parameter must be valid UTF-8
• Overwrites file with UTF-8 encoding

Bash Tool

• Set locale before commands: LANG=en_US.UTF-8 command
• Pipe through iconv if needed: command | iconv -f GBK -t UTF-8