Skills Janitor

9 个命令,可审核, 去重, 检查, 修复并跟踪您的 Claude 代码技能。它会显示哪些代码有问题, 哪些代码重叠以及哪些代码您从未修改过。免费,open-source 。

9 commands to review, deduplicate, check, fix, and track your Claude code skills. It shows which code has issues, which code overlaps, and which code you have never modified. Free, open source.

In-depth Report

  • Skills Janitor is an open source skills management plug-in built specifically for Claude Code users, released in April 2026 by developer Krzysztof Hendzel. It uses 9 independent slash commands to help users audit, deduplicate, verify, repair and track local skill libraries, solving the common pain point of "installing dozens of skills but not being able to name them." The product is open source under the MIT license and has zero external dependencies. It received initial approval from 149 likes and 6 comments on Product Hunt.

  • Skills Janitor was born out of real troubles experienced by developers themselves. The creator shared in the Reddit community: "I installed 35 skills, but I can't even name 10 of them." This scene accurately hits the pain point of many Claude Code users - with the booming skills market (for example, Claudate claims to have 150,000+ skills), users' local skill directories gradually fall into a state of confusion, duplication, and forgetfulness. The product was officially launched on Product Hunt on April 13, 2026, and the GitHub warehouse received 30 stars and 4 forks. As a pure shell script implementation tool, it only requires Bash, Python 3 and curl to run, without external dependencies such as pip or node modules. This minimalist design greatly lowers the threshold for use.

  • The core value of Skills Janitor is to transform skills management, an originally fuzzy, manual process, into a structured, automated workflow. It provides 9 independent slash commands, each of which is a self-contained skill module: Audit and inventory class: /janitor-audit can list a complete list of all installed skills; /janitor-usage helps users identify "dead weight" by parsing the conversation history and displaying skills that are actually called and never used. A typical case is shown in the GitHub README: a user installed 36 skills in 4 weeks, but only 4 were actually actively used. Quality and repair categories: /janitor-check scans skill files for errors and missing metadata; /janitor-fix provides automatic repair functions and previews changes in dry running mode by default to ensure safety; /janitor-cleanup is responsible for finding and removing damaged or invalid skills. Optimization and discovery categories: /janitor-duplicates identifies duplicate skills with the same functionality; /janitor-search supports searching for new Claude Code skills by keyword on GitHub; /janitor-compare compares local skills with similar alternatives on GitHub to assist decision-making. Report generation: /janitor-report generates complete health check reports for all skills to facilitate regular maintenance. It is worth mentioning that all commands support natural language triggering. Users can directly enter descriptive statements such as "audit my skills" and "what skills I use", and the system will automatically call the corresponding functions through keyword matching. This design reduces memory costs and is in line with Claude Code's own interaction habits.

  • Skills Janitor adopts a completely open source and free strategy and is released based on the MIT license. Users can install it in two ways: one is to install it directly through Claude Code's plug-in market, and the other is to clone it from GitHub to the local skills directory. This zero-cost, low-barrier positioning allowed it to quickly gain attention in the developer community. From a business model perspective, Skills Janitor itself does not directly generate revenue. Its potential value lies in: establishing developer influence as a traffic portal; accumulating technical reputation through open source contributions; and laying the foundation for possible future value-added services (such as enterprise version, cloud synchronization function). Currently, similar tools in the Claude Code ecosystem mostly adopt a hybrid model of open source + donation or open source + enterprise support.

  • Due to the short launch time of the product (about 1 week), public user reviews are relatively limited, but some consensus can be extracted from existing feedback. In terms of positive reviews, users of the PH Chinese website said that "I finally don't have to check manually, which saves time" and "the duplicate removal function is very practical, and I found several repeatedly defined skills." The open source attribute has also been recognized by the community. Some users commented "Long live open source, a must-have artifact for developers." The security design has been well received - mechanisms such as default dry operation, deletion requiring confirmation, and no modification of market skills allow users to dare to try it in a formal environment. In terms of improvement suggestions, some users expressed the hope that the graphical report export function could be added, implying that the current pure command line output form is not friendly enough to non-technical users. In addition, as a new tool, its long-term stability and ability to handle edge cases have yet to be tested over time.

  • The Claude Code skill management track where Skills Janitor is located is a subdivision of the entire AI coding tool ecosystem. According to the awesome-claude-code warehouse, there are currently dozens of related tools in the ecosystem, covering skills markets (such as Claudate, SkillsMP), skill creation tools (such as Skill-Creator), usage monitoring (such as ccflare, ccxray) and other directions. In the segment of skills management, Skills Janitor is positioned more as a "maintenance tool" rather than a "creation tool". It is differentiated from competing products such as AI Skills Manager - the latter focuses more on the full life cycle management of skills (creation, editing, synchronization), while Skills Janitor focuses on the auditing, cleaning and optimization of existing skills. This positioning is similar to the "system cleaning tools" in the traditional software ecosystem, which solves the problem of "skill expansion" that has become increasingly prominent as the ecosystem matures. Industry observers believe that with the expansion of Claude Code users and the prosperity of the skills market, skills management tools will evolve from "niche needs" to "rigid needs." Skills Janitor’s early layout gives it the opportunity to become a standard tool in this segment.

  • As a newly released open source tool, Skills Janitor has yet to experience any significant controversy or negative incidents. However, from the perspective of product form and ecological position, potential risks include: Functional boundary risk: The core function of Skills Janitor is analysis and suggestions, not forced cleanup. It respects the design that "some overlap is intentional" and doesn't automatically delete anything. Although this conservative strategy is safe, it also means that users still need to manually perform the final decision, and the degree of automation is limited. Ecological dependence risk: Skills Janitor completely relies on Claude Code’s skill system. If Anthropic adjusts the skill format or introduces native skill management functionality in the future, the applicability of this tool may be affected. Competitive substitution risk: There are already multiple skills management related tools in the Claude Code ecosystem, including the more comprehensive AI Skills Manager, the more integrated claude-skills-manager, etc. Skills Janitor needs to find a balance between "ease of use" and "full functionality" to remain competitive.

  • Suitable for use by: Developers who have installed a lot of Claude Code skills but lack system management Efficiency seekers who want to regularly clean up their skill library and keep their workspace streamlined Team technical leader who needs to audit skill quality and troubleshoot configuration errors Early adopters who are friendly to open source tools and willing to try new tools Not suitable for people: Light users who use only a few built-in skills and have no customization needs Those who want "one-click cleaning" that is fully automated and requires no manual intervention Users who are unfamiliar with the command line interface and prefer graphical operations Usage recommendations: It is recommended that Skills Janitor be included in regular maintenance processes, such as running /janitor-usage and /janitor-report once a month to identify long-term unused skills and potential issues. For team environments, consider integrating skills health checks into your CI/CD process to ensure that all members’ skill libraries remain consistent.

  • Skills Janitor is a practical tool that accurately solves the pain points of Claude Code users. It does not try to be a "Swiss Army Knife" with large functions, but focuses on the single scenario of "skills auditing and cleaning". It provides users with an out-of-the-box solution through 9 concise commands and a zero-dependency design. In the context of the rapid expansion of the Claude Code ecosystem, the value of such maintenance tools will become increasingly prominent. For developers who have been troubled by "skill bloat", Skills Janitor is worth a try.

User Reviews

  • 头像
    MichelleBennettJr
    终于有人做了这个,之前手动查重复累死了

  • 头像
    LineMichel
    装了一堆skill根本用不过来,用了skills-janitor才发现35个里只有4个是真正在用的,太真实了

  • 头像
    MRogers_7
    去重功能太实用了,我的skills目录简直是个垃圾堆

  • 头像
    Marilyn.Kim_7
    零依赖好评

  • 头像
    淡然370
    开源免费还要什么自行车

  • 头像
    浮生749
    dry-run模式好评,不用担心误删

  • 头像
    AlissaFrancois
    感觉对新手不太友好,命令行还是有点门槛

  • 头像
    Thomas.Jackson_88
    说白了就是给Claude Code的技能库减肥

  • 头像
    Rebecca.Cooper_2022
    推荐集成到CI/CD里定期跑

  • 头像
    ZStewart_2024
    用了这个才知道自己装了那么多重复的skill

  • 头像
    JHoward_2021870
    希望能出个图形界面

  • 头像
    NatalieMoore_20213
    我的skill目录终于清爽了

  • 头像
    Donald_Lee_99
    比手动清理省事多了

  • 头像
    RSanchezK
    无遥测这个要点赞

  • 头像
    MsBobHenry_x
    开发者必备

  • 头像
    RMurphy_2020625
    简单直接,不错

  • 头像
    DavidMurrayK
    audit功能挺准的,一下就扫出来8个坏掉的skill

  • 头像
    Natalie_Rivera_2022
    建议增加导出PDF报告功能

  • 头像
    brownwolf389
    MIT协议就是香

  • 头像
    徐琪婷
    如果只装了几个skill可能没必要装这个

  • 头像
    HHarris007
    给团队用很合适

  • 头像
    EAdamsX
    找重复定义超好用

  • 头像
    DrBradleySutton_2024
    YYDS

  • 头像
    SandyFreier
    解决了实际问题

  • 头像
    JulieStewart_2020
    清理完感觉整个人都清爽了哈哈哈

  • 头像
    Kathryn_RobertsJr097
    期待后续更新

  • 头像
    goldenkoala825
    唯一缺点就是不能删完之后自动帮我补上真正需要的skill哈哈

  • 头像
    WReyes_66812
    安装简单,配置也简单

  • 头像
    MJonesJr96
    已经加入日常开发流程了