AgentSkillsCN

Brew Optimize

分析并优化 Homebrew 安装——查找未使用、过时或孤立的软件包（仅限 macOS）

SKILL.md
--- frontmatter
description: Analyze and optimize Homebrew installations - find unused, outdated, and orphaned packages (macOS only)
user-invocable: true
allowed-tools:
  - Bash
  - AskUserQuestion

Perform a comprehensive Homebrew optimization audit. Run all analysis steps, then present a summary with actionable recommendations. Ask before making any changes.

Step 1: Gather Data

Run all of these in parallel:

  • •brew leaves (top-level formulae)
  • •brew list --cask (installed casks)
  • •brew outdated (outdated packages)
  • •brew autoremove --dry-run (orphaned dependencies)
  • •brew cleanup --dry-run (stale downloads/old versions)
  • •brew doctor 2>&1 (general health warnings)

Step 2: Disk Usage

Find the top 10 largest packages by cellar size:

code
du -d1 $(brew --cellar) 2>/dev/null | sort -rn | head -11

Also report total cellar size: du -sh $(brew --cellar)

Step 3: Analyze Usage

For each top-level formula from brew leaves:

  • •Get the package prefix: brew --prefix <pkg>
  • •Check last access time of binaries via stat -f "%Sa" -t "%Y-%m-%d" on files in the bin/ directory
  • •For library-only packages (no bin/), check files in lib/ or include/ instead
  • •Flag packages not accessed in 3+ months as "stale"
  • •Flag packages not accessed in 6+ months as "likely unused"

For each cask:

  • •Check last opened date via mdls -name kMDItemLastUsedDate on the .app bundle
  • •Flag apps not opened in 3+ months

Step 4: Find Orphaned Dependencies

For each outdated package, run brew uses --installed <pkg>:

  • •If no dependents: mark as "orphaned, safe to remove"
  • •If has dependents: list what depends on it

Step 5: Detect Duplicate Versions

Check for multiple versions of the same tool:

  • •brew list --formula | grep -E '@[0-9]' to find versioned formulae
  • •Group by base name (e.g., python@3.11 + python@3.13, llvm@19 + llvm@20)
  • •For each group, check which versions have dependents via brew uses --installed
  • •Flag versions with no dependents as removal candidates

Step 6: Present Report

Organize findings into these categories:

  1. •Likely Unused - top-level packages with stale access times and no dependents
  2. •Orphaned Dependencies - outdated packages nothing depends on
  3. •Duplicate Versions - multiple versions of same tool, with dependent info
  4. •Outdated (in use) - packages that are outdated but actively depended upon
  5. •Stale Casks - cask apps not recently opened
  6. •Cleanup Savings - space reclaimable via brew cleanup
  7. •Doctor Warnings - issues reported by brew doctor (if any)

For each category, show a markdown table with package name, relevant details (last access, version, size, dependents).

End with concrete commands the user can copy-paste:

  • •brew uninstall ... for removals
  • •brew upgrade ... for updates
  • •brew cleanup for cache cleanup

Ask the user which actions they want to take before executing anything.