Custom hooks and skills for Claude Code.
A Claude Code hook that supercharges git worktree workflows with Ghostty.
When you run /worktree in Claude Code, this hook:
- Creates a git worktree in a sibling directory (not inside the repo)
- Prompts for a custom branch name via macOS dialog
- Opens a Ghostty 2x2 split layout:
+---------------+---------------+
| Claude Code | lazygit |
| (top-left) | (top-right) |
+---------------+---------------+
| Terminal | yazi |
| (bottom-left) | (bottom-right)|
+---------------+---------------+
On /worktree removal, it cleans up the worktree, branch, and empty directories.
Requirements: jq, Ghostty, lazygit, yazi
Ghostty keybindings needed:
super+d = new_split:right
super+shift+d = new_split:down
super+alt+left = goto_split:left
Installation:
# Copy the hook
mkdir -p ~/.claude/hooks
cp hooks/worktree-ghostty.sh ~/.claude/hooks/
# Add to ~/.claude/settings.jsonAdd this to your ~/.claude/settings.json:
{
"hooks": {
"WorktreeCreate": [
{
"hooks": [
{
"type": "command",
"command": "bash ~/.claude/hooks/worktree-ghostty.sh",
"timeout": 30
}
]
}
],
"WorktreeRemove": [
{
"hooks": [
{
"type": "command",
"command": "bash ~/.claude/hooks/worktree-ghostty.sh",
"timeout": 15
}
]
}
]
}
}A /humanize skill that strips AI fingerprints from text. Based on Wikipedia's "Signs of AI writing" guide.
Detects and rewrites 13 pattern categories:
- Inflated symbolism & significance
- Promotional language
- Superficial -ing analysis
- Vague attribution (weasel words)
- Em dash overuse
- Rule of three
- AI vocabulary words
- Negative parallelism
- Excessive conjunctive phrases
- Editorializing & sycophantic tone
- Formatting tells
- Copula avoidance & synonym cycling
- False ranges & compulsive summaries
Installation:
# Symlink to your Claude skills directory
mkdir -p ~/.claude/skills
ln -s "$(pwd)/skills/humanize" ~/.claude/skills/humanizeThen use /humanize in any Claude Code session, or just say "humanize this text".
MIT