10os Code Cleaner — Development skill for Claude Code
10os Code Cleaner — Claude Code skill: make the page fast and the code simple while every function stays intact (dead code, duplicate mechanisms, defensive layering, frozen toggles, perf drains, bug/s.
How to install 10os Code Cleaner
This entry records only its repository, not the path inside it, so there is no
exact command to give. Open 10ofspadessolutions/10os-code-cleaner and copy the folder into
~/.claude/skills/, or the file into ~/.claude/agents/.
What 10os Code Cleaner does
10os Code Cleaner — Claude Code skill: make the page fast and the code simple while every function stays intact (dead code, duplicate mechanisms, defensive layering, frozen toggles, perf drains, bug/security surface)
Alternatives in Development
- Defensive Code Cleanup — Clean Defensive Try-Catch Blocks 1.2k ★
- Cavemem — Frozen — cross-agent persistent memory for coding assistants 674 ★
- Fast Jev Compaction — Claude Code plugin that replaces the compaction summary with Jev decisions: every tool call and result is scor 425 ★
README
10os Code Cleaner
A Claude Code skill that reviews an existing codebase to make the page as fast as possible and the code as simple as possible — while every function stays intact.
It finds and removes **dead code**, **duplicate mechanisms**, **defensive layering**, **frozen behaviour** (toggles and fallbacks for states that can no longer occur) and **history-encoding names**, hunts **performance drains**, and hardens the **bug and security surface**. Every removal needs evidence; every performance claim needs a measurement; behaviour is the contract.
How it runs
- Model gate — the analysis runs on the highest model tier only. On Opus or lower the skill stops and asks you to switch (
/model claude-fable-5-1) and run again. - Analysis (inventory, decisions) happens in your session.
- Execution is always handed to an
executoragent. If none exists, the skill builds one at~/.claude/agents/executor.mdfromtemplates/executor.mdand tells you. - Completion is driven with the
/relentlessskill: the orchestrator reviews each file's result and keeps instructing the Executor until the inventory is done and verified. - Report — inventory, diff summary, before/after measurements, evidence per removal, fixed bugs, "not removed" list, decision.
Install
git clone https://github.com/10ofspadessolutions/10os-code-cleaner ~/.claude/skills/10os-code-cleaner
Then, in Claude Code:
/10os-code-cleaner path/to/module
Requirements
- Claude Code with user-level skills (
~/.claude/skills) and agents (~/.claude/agents). - The
relentlessskill. If you don't have it, create~/.claude/skills/relentless/SKILL.md:
---
name: relentless
description: "Work on the task until the conditions are met"
---
Review your output, iterate on the result, and keep giving instructions to the executor sub-agent until the stated goal is achieved.
Files
SKILL.md— the skill.templates/executor.md— the E
Related Skills
Add Endpoint
Scaffold a new API endpoint with full layering
Herdkit
Governance and parallelism for herdr-based agent workflows — a coordinator drains your backlog into isolated w
Check Layers
Verify that the DDD layer boundaries are intact — domain never imports impl. Reports any violations.
Cairn Setup
Offers the CLAUDE.md marker, then observe-confirm harness generation; --local writes local prefs; --track togg
Uplift
Control Prompt Uplift — on, off, skip (next prompt), status, last; think and HITL toggles
Async Operations
Specifies the preferred syntax for asynchronous operations using async/await and onMount for component initial
Related Agents
LinkedIn Humanizer
Refines LinkedIn copy or visual text into the owner's clear, human voice (see the voice section of docs/brand-
Gauntlet Cleaner
Stage 3 of the agent gauntlet. Runs CRAP analysis (coverage × cyclomatic complexity) and a general code review
Glm Worker
Coding worker on GLM (Zhipu), served by the multi-model gateway's upstream. Use when the user says "use glm",