Farty Bobo Installer

Fetch the live catalog from fartybobo/farty-bobo, let the user pick what they want, and install it to the right places on this machine.

1. Fetch the catalog

Run all three in parallel:

# Skills
gh api "repos/fartybobo/farty-bobo/contents/skills" --jq '.[].name'

# Hooks (.sh files only)
gh api "repos/fartybobo/farty-bobo/contents/hooks" \
  --jq '[.[] | select(.name | endswith(".sh")) | .name]'

# Commands
gh api "repos/fartybobo/farty-bobo/contents/commands" --jq '.[].name'

For each skill, fetch its description from the frontmatter:

gh api "repos/fartybobo/farty-bobo/contents/skills/<name>/SKILL.md" \
  --jq '.content' | base64 -d | head -6

For each hook, fetch its purpose from the file header:

gh api "repos/fartybobo/farty-bobo/contents/hooks/<name>" \
  --jq '.content' | base64 -d | head -5

Fetch all descriptions in parallel — do not loop sequentially.

2. Present the menu

Display a categorized list:

Available from fartybobo/farty-bobo:

SKILLS
  • <name> — <description from frontmatter>
  ...

HOOKS
  • <name> — <purpose from header comment>
  ...

COMMANDS
  • <name>
  ...

Ask the user:

“Which would you like to install? Type ‘all’, a category name (skills / hooks / commands), or a space-separated list of names.”

Parse the response. If ambiguous, ask for clarification.

3. Download selected items

Skills

For each selected skill, list the files in its directory:

gh api "repos/fartybobo/farty-bobo/contents/skills/<name>" --jq '.[].name'

For each file in the skill directory, download it:

gh api "repos/fartybobo/farty-bobo/contents/skills/<name>/<file>" \
  --jq '.content' | base64 -d > ~/.claude/skills/<name>/<file>

Create ~/.claude/skills/<name>/ if it doesn’t exist. If the skill directory already exists, ask the user: “~/.claude/skills// already exists — overwrite, skip, or merge?"

Hooks

For each selected hook, download the script:

gh api "repos/fartybobo/farty-bobo/contents/hooks/<name>" \
  --jq '.content' | base64 -d > ~/.claude/hooks/<name>
chmod +x ~/.claude/hooks/<name>

Create ~/.claude/hooks/ if it doesn’t exist.

If a hook file already exists, ask: “~/.claude/hooks/ already exists — overwrite or skip?"

After downloading, go to Step 4 (hook registration) for each hook installed.

Commands

For each selected command, download it:

gh api "repos/fartybobo/farty-bobo/contents/commands/<name>" \
  --jq '.content' | base64 -d > ~/.claude/commands/<name>
chmod +x ~/.claude/commands/<name>

Create ~/.claude/commands/ if it doesn’t exist.

4. Register hooks in settings.json

This step only applies if one or more hooks were installed.

First, fetch the hook registration config from the source repo’s settings.json to know exactly how each hook should be wired up (event type, matcher, command path):

gh api "repos/fartybobo/farty-bobo/contents/settings.json" \
  --jq '.content' | base64 -d

Extract the hooks entries that correspond to each installed hook script.

Then ask the user:

“Where should the hooks be registered?

1. Globally — ~/.claude/settings.json (fires in every project)
2. This project only — .claude/settings.json in the current directory”

Based on their choice, read the target settings.json (create it with {} if it doesn’t exist), merge in the new hook entries under the correct event keys (PreToolUse, PostToolUse, SessionStart, etc.), and write it back. Do not overwrite existing hook entries for other scripts — append only.

If a hook entry for the same script path already exists in the target settings.json, skip it and tell the user it was already registered.

5. Report

Print a summary:

Installed:
  ✓ skills/  <name>, <name>, ...
  ✓ hooks/   <name> (registered in ~/.claude/settings.json)
  ✓ commands/ <name>, ...

Skipped (already existed):
  - <name>

Done. Restart Claude Code for hook changes to take effect.