fix-nerd-fonts

Fix Nerd Fonts

Nerd Font icons use Unicode Private Use Area codepoints (U+E000-U+F8FF and U+F0000-U+10FFFF). The Write and Edit tools may silently strip or corrupt these characters. This skill ensures they survive every edit.

When This Skill Applies

After every Write or Edit to a file that contains Nerd Font icons. Currently the known files are:

File	Icons
tui/src/main.rs	nf-md-web (U+F059F), nf-fa-keyboard_o (U+F11C), nf-fa-user (U+F007)
issues/0504-web-tui/README.md	nf-md-web (U+F059F), nf-fa-keyboard_o (U+F11C), nf-fa-user (U+F007), nf-md-refresh (U+F0450)

Update this table when new files or icons are added.

The Problem

The Write and Edit tools transmit file content as text. Characters in the Private Use Area — especially Supplementary Private Use Area codepoints above U+FFFF — may be silently dropped, replaced with replacement characters, or truncated. The file saves without error, but the icons are gone.

This is invisible in diffs and code review because the missing character leaves no trace — just an empty string where the icon was.

Step 1: Verify Icons

After any edit, run the verification script:

python3 -c "
src = open('<file_path>').read()
icons = {
    'nf-md-web (U+F059F)': '\U000F059F',
    'nf-fa-keyboard_o (U+F11C)': '\uF11C',
    'nf-fa-user (U+F007)': '\uF007',
    'nf-md-refresh (U+F0450)': '\U000F0450',
}
for name, char in icons.items():
    if char in src:
        print(f'  OK  {name}')
    else:
        print(f'  MISSING  {name}')
"

Adjust the icon list to match what the file should contain.

Step 2: Fix Missing Icons

If any icon is missing, use the placeholder-and-replace pattern:

1. In the Edit tool, use a unique ASCII placeholder string where the icon should go (e.g., PLACEHOLDER_WEB, PLACEHOLDER_KEYBOARD).
2. Then run Python to replace the placeholder with the real Unicode character:

python3 -c "
src = open('<file_path>').read()
src = src.replace('PLACEHOLDER_WEB', '\U000F059F')
src = src.replace('PLACEHOLDER_KEYBOARD', '\uF11C')
src = src.replace('PLACEHOLDER_USER', '\uF007')
src = src.replace('PLACEHOLDER_REFRESH', '\U000F0450')
open('<file_path>', 'w').write(src)
"

Step 3: Verify Again

Re-run the verification script from Step 1 to confirm all icons are present.

Python Escape Syntax

Codepoints at or below U+FFFF use \uXXXX:

'\uF11C'   # U+F11C  (nf-fa-keyboard_o)
'\uF007'   # U+F007  (nf-fa-user)

Codepoints above U+FFFF use \U00XXXXXX (8 hex digits, zero-padded):

'\U000F059F'  # U+F059F  (nf-md-web)
'\U000F0450'  # U+F0450  (nf-md-refresh)

Using \uF059F for a codepoint above U+FFFF is wrong — Python interprets it as \uF059 + literal F, producing a garbage character.

Rule: Always Use Unicode Escape Syntax in Source Code

NEVER embed raw Nerd Font bytes in source code. Always use the language's Unicode escape syntax. Raw UTF-8 bytes are silently corrupted by text editors, LLM tools, clipboard operations, and diff/patch workflows. Unicode escapes are pure ASCII and survive any tool chain.

Language	Syntax	Example (U+F007)	Example (U+F059F)
Rust	\u{XXXX}	"\u{F007}"	"\u{F059F}"
Zig	\u{XXXX}	"\u{F007}"	"\u{F059F}"
Python	\uXXXX / \U00XXXXXX	'\uF007'	'\U000F059F'
C/C++	\uXXXX / \UXXXXXXXX	u8"\uF007"	u8"\U000F059F"
Swift	\u{XXXX}	"\u{F007}"	"\u{F059F}"
JavaScript	\u{XXXX}	"\u{F007}"	"\u{F059F}"

When editing a file that contains raw Nerd Font bytes, convert them to Unicode escapes as part of the edit. Use xxd or python3 to identify the codepoint, then replace the raw bytes with the appropriate escape.

Adding New Icons

When introducing a new Nerd Font icon to the project:

1. Add the icon name, codepoint, and file to the table in this skill.
2. Add it to the verification script's icons dict.
3. Add a placeholder constant for it.
4. Use the placeholder-and-replace pattern for the first embed.
5. Use Unicode escape syntax (\u{...} in Rust/Zig, etc.) — never raw bytes.