claude-code-tuneup

Audit a Claude Code installation for duplicate skills, broken hooks, and context bloat.

2|Updated Jul 28, 2026
One-click install
npx skills add https://github.com/mbadali25/useful-claude-add-ons --skill claude-code-tuneup-mbadali25
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: claude-code-tuneup
Source: https://github.com/mbadali25/useful-claude-add-ons/tree/main/skills/claude-code-tuneup
Command: npx skills add https://github.com/mbadali25/useful-claude-add-ons --skill claude-code-tuneup-mbadali25

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) components.

What problem does it solve? Claude Code setups accumulate duplicate skills, broken hooks, disabled plugins, and oversized CLAUDE.md files that make sessions slow to start, laggy on every tool call, and prone to early compaction. This Skill measures those problems from disk and produces a ranked cleanup plan with the exact fix command per finding. ## Core Features & Use Cases - Read-only audit script: Runs a stdlib-only Python script that inventories settings files, loose and plugin skills, hooks, subagents, MCP servers, marketplaces, and memory files without modifying anything. - Ranked findings with fixes: Reports HIGH/MED/LOW findings such as duplicate skill installs, hooks referencing missing scripts, per-Bash-call hook latency, and oversized context preloads, each with the exact remediation command. - Context budget table: Estimates the tokens loaded before the first message from skill descriptions, CLAUDE.md, rules files, agents, and MCP tool schemas. - Use Case: After running an install script that added dozens of plugins and marketplaces, a user finds Claude Code sluggish. The audit reveals a skill installed both loosely and via plugin, three PreToolUse hooks shelling out to Git Bash, and a disabled plugin still being refreshed, then proposes safe removals. ## Quick Start Ask the assistant to audit your Claude Code setup for what is making it slow and produce a ranked cleanup plan.

Frequently Asked Questions about claude-code-tuneup

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I find out why Claude Code is slow to start?

Run the cc_audit.py script, which groups hooks by event and counts registered marketplaces. SessionStart hooks spawn processes and inject context at every session start, and each marketplace is a git fetch on refresh, so both directly cause startup latency.

How do I fix a skill that appears twice in Claude Code?

Duplicate skills usually mean a loose copy in ~/.claude/skills/ shadows an enabled plugin copy. Delete the loose copy after confirming with diff -r that it contains no local edits, since the plugin copy is the one that receives updates.

Does the audit script modify or delete anything?

No, cc_audit.py is strictly read-only and stdlib-only Python 3.8+. It reports findings with the exact fix command for each, and the skill requires explicit user approval before any deletion, uninstall, or settings edit.

Why does Claude Code compact too early or run out of context?

Two causes: preloaded content like CLAUDE.md, rules files, skill descriptions, agent listings, and MCP tool schemas filling the window, or CLAUDE_AUTOCOMPACT_PCT_OVERRIDE set below 70. Confirm the real breakdown with /context in a live session.

Why does every Bash command feel laggy in Claude Code?

PreToolUse and PostToolUse hooks run serially around every matching tool call, so several hooks on the Bash matcher add their combined latency to each command. On Windows, hooks shelling out to bash.exe add roughly 100-300 ms of Git Bash startup per call.

When should I disable versus uninstall a Claude Code plugin?

Disable when unsure because it is reversible in one command; uninstall when done with it, because a disabled plugin still has its marketplace refreshed and occupies disk. Never remove a marketplace that still has an installed plugin.