Cleanup Unused Files

Optimize for reclaiming disk space fast.

Prefer obvious, reproducible junk over careful semantic analysis.

Default stance

• Target caches, build output, temp folders, and local tool state first.
• Prefer exact paths over broad globs.
• Be explicit about blast radius.
• Treat anything under source directories as higher risk unless it is known build output.
• If the cleanup touches machine-wide tool state or login/session data, call that out clearly before running it.

Primary targets

Common high-yield cleanup targets include:

• project build output like target/, dist/, .next/, coverage/, .turbo/
• package-manager caches
• SDK/tool caches
• app caches under ~/Library/Caches
• stale temp folders
• old logs, dumps, and exports
• backup/swap/temp files

Examples of the user's preferred cleanup style:

rm -rf ~/.codex
rm -rf ~/Library/Caches
rm -rf ~/.rustup
rm -rf ~/.bun
rm -rf ~/.npm
rm -rf ~/.lmstudio
rm -rf src-tauri/target

Also valid tool-native cleanup commands when they fit better:

cargo clean --manifest-path src-tauri/Cargo.toml
npm cache clean --force
brew cleanup --prune=all
rustup toolchain list | grep -v default | xargs -I{} rustup toolchain uninstall {}

Workflow

1) Identify cleanup scope

Determine whether the user wants:

• project-only cleanup
• home-directory cache cleanup
• machine-wide developer reset
• maximum disk recovery regardless of re-download cost

If the request is vague, default to an audit of likely large cache directories first.

2) Prefer highest-yield directories

When disk recovery is the goal, focus on these before hunting tiny files:

• ~/Library/Caches
• ~/.npm
• ~/.bun
• ~/.rustup
• ~/.cargo
• ~/.codex
• ~/.lmstudio
• repo-local target/, dist/, .next/, coverage/
• large temp/export folders under the workspace or home directory

3) Distinguish safe nukes from state resets

Usually safe and reproducible:

• build output
• download caches
• derived data
• package caches
• logs
• temp files

Higher impact, but still valid if explicitly requested:

• ~/.rustup (removes installed toolchains)
• ~/.cargo (removes installed cargo binaries, registry, and config-adjacent state)
• ~/.codex
• ~/.bun
• ~/.npm
• ~/.lmstudio
• all of ~/Library/Caches

Call out that these may force reinstallation, re-downloads, reindexing, or loss of local session/tool state.

4) Execute with exact commands

Prefer short, direct commands with obvious targets.

Examples:

rm -rf ~/.codex
rm -rf ~/Library/Caches
rm -rf ~/.rustup
rm -rf ~/.bun
rm -rf ~/.npm
rm -rf ~/.lmstudio
rm -rf src-tauri/target

For repo build cleanup, exact-path deletion is fine.

For machine-wide cleanup, group commands by tool/system and explain briefly what each group resets.

Avoid fancy one-liners if they make the blast radius less legible.

5) Summarize outcome

Report:

• what paths were removed
• which ones were skipped
• likely consequences such as re-downloads or tool re-bootstrap
• any follow-up commands the user may need

Decision rules

Delete immediately when:

• the user explicitly names the path
• the directory is clearly cache/build output/tool state
• the directory is reproducible and not the source of truth

Pause and confirm when:

• the path is broad and not explicitly requested
• the path may contain user-authored content
• the cleanup may remove credentials, custom models, local databases, or app data beyond cache
• the command would recurse through mixed-purpose directories

Preferred style

Be blunt and practical.

When asked for a cleanup recipe, produce something like this:

# Rust / Tauri build output
rm -rf src-tauri/target

# Tool and package caches
rm -rf ~/.npm
rm -rf ~/.bun
rm -rf ~/.codex
rm -rf ~/.lmstudio

# macOS app caches
rm -rf ~/Library/Caches

# Rust toolchains and state
rm -rf ~/.rustup

Favor the biggest wins first.