claude-md-management for Claude Code: review, how it works and whether it is worth it (2026)

What you get in one minute#

claude-md-management is a free Anthropic plugin with one skill and one command that audit your CLAUDE.md files and add what a session taught Claude; it is small and safe to read, but built-in commands now cover much of it.

Logo: the official Claude Code logo from Anthropic's press kit (anthropic.com/press-kit)

The marketplace page shows "287,247" installs (October 2026). It has known gaps with AGENTS.md, rules files and the local file name (see below).

What it does#

You ask Claude to audit your CLAUDE.md, or you type a command at the end of a session, and Claude proposes small edits that you approve one by one.

Added at version 1.0.0:

• 0 hooks, 0 agents, 0 MCP servers, 0 scripts. Nothing runs by itself.

How it works in the background#

Both pieces are Markdown instructions that tell Claude to read, report, ask and only then edit. The picture follows each path.

• Always on: only the skill name and description (338 characters) and the command description.
• On use: the skill body, then the three reference files as Claude reads them.
• Approval: both paths say to ask before writing; nothing enforces it except Claude Code's own edit prompt.

Under the hood: dissection#

The plugin is 7 small text files (794 lines, 18 KB) plus two screenshots and a licence; there is no code that runs.

The file tree#

claude-md-management/
├── .claude-plugin/plugin.json      name, version 1.0.0, author Anthropic
├── README.md                       two usage examples
├── LICENSE                         Apache 2.0
├── claude-md-improver-example.png  README screenshot (519 KB)
├── revise-claude-md-example.png    README screenshot (556 KB)
├── commands/
│   └── revise-claude-md.md         the command prompt, 54 lines
└── skills/claude-md-improver/
    ├── SKILL.md                    the 5-phase audit workflow, 179 lines
    └── references/
        ├── quality-criteria.md     the scoring rubric, 109 lines
        ├── templates.md            CLAUDE.md templates by project type, 253 lines
        └── update-guidelines.md    what to add and what not, 150 lines

What each file is#

Key excerpts#

The rubric weights, from quality-criteria.md: commands 20, architecture 20, patterns 15, conciseness 15, currency 15, actionability 15.

The discovery step, from SKILL.md:

find . -name "CLAUDE.md" -o -name ".claude.md" -o -name ".claude.local.md" 2>/dev/null | head -50

The line the security issues point at, from the same rubric file:

Run documented commands (mentally or actually)

Disk and network#

• Reads: every CLAUDE.md-style file under the folder you start in, plus your code to check it.
• Writes: only through Claude's Edit tool, after you approve a diff.
• Network and dependencies: none in the plugin; it uses Claude Code's own Read, Glob, Grep, Bash and Edit.

Chart: file sizes from the GitHub contents of the plugin folder at main (commit 41f619d), 27 Oct 2026. Only the loaded parts count against your context.

What users say#

13 user reports read: 2 praise, 2 mixed, 10 complaints, but the split is skewed. Complaints are mostly GitHub issues in the plugin's own repository, which attract problems. Hacker News had no hits for the plugin name, and Reddit was not read.

Chart: our count of the 13 reports, October 2026. Not a satisfaction rate.

Chart: themes of the 10 complaints, labelled by DIY Lab Hub, October 2026.

Pros: why it is useful#

It turns "keep CLAUDE.md healthy" into a checklist you can run, and it never edits without a diff.

• A clear rubric: six criteria and a grade show what is missing.
• Matches the docs: "one line per concept" fits the official advice to stay under 200 lines.
• Low risk to read: no hooks, scripts or network; about 18 KB of text.
• Free and official: from Anthropic, Apache 2.0, with 287,247 listed installs.

Cons: why it is not useful#

Much of it is now built in, and it lags the current memory system.

• Overlap: the docs describe /init (suggests improvements to an existing CLAUDE.md), /memory and /doctor prompt-audit (v2.1.283+), which flags stale commands and contradictions.
• Misses files: no AGENTS.md, .claude/rules/, .claude/CLAUDE.md or @ imports; checked in the file text on main.
• Wrong local file: advice to put personal notes in .claude.local.md is silently ignored by Claude Code (per the docs).
• Stale tip: the skill recommends a # shortcut; we found no mention of it in the current memory docs (not tested).
• Open bugs: the eight issues linked here were still open on 27 Oct 2026.

Should you consider it? When it is useful#

Consider it only if you maintain several CLAUDE.md files and want a scored audit; otherwise use the built-in commands.

Install and check it#

You install it from Anthropic's official marketplace with one command, inside Claude Code or from your shell.

Linux#

Inside Claude Code:

/plugin install claude-md-management@claude-plugins-official

Or from your shell:

claude plugin install claude-md-management@claude-plugins-official

Windows (PowerShell)#

The plugin has no start-up hook, so it needs no Git Bash to load.

claude plugin install claude-md-management@claude-plugins-official

Mac (Terminal)#

claude plugin install claude-md-management@claude-plugins-official

Check it worked#

claude plugin list
claude plugin details claude-md-management

The second should show one skill and one command plus the token cost. Then type audit my CLAUDE.md files; Claude should print a "CLAUDE.md Quality Report" before it edits anything.

Things to keep in mind#

• Security. The plugin has no code, but it tells Claude to run find and, per the rubric, documented commands. Audit only repositories you trust, and do not let it run commands you did not read.
• Prompt injection. Third-party CLAUDE.md files (for example in node_modules) are read as instructions-like text.
• Updates. The version has stayed 1.0.0; the docs say a set version keeps you on it until it changes, so fixes may reach you late.
• Your files change. Work in git, so any approved edit can be undone.

Troubleshooting#

FAQ#

Is it made by Anthropic? Yes; plugin.json names Anthropic and the marketplace page says "Anthropic verified".

Does it change my files without asking? Both prompts say to show diffs and ask first; Claude Code's edit permission is a second check.

Does it replace /init? No. /init creates a CLAUDE.md; this plugin scores and updates existing ones.

Sources#

• anthropics/claude-plugins-official, plugins/claude-md-management: every file and the marketplace entry, read at commit 41f619d (27 October 2026).
• claude-md-management on the Claude marketplace: install count and components (October 2026).
• How Claude remembers your project, Skills and the plugin manifest reference: official Claude Code docs (October 2026).
• User reports: GitHub issues 273, 490, 1751, 2790, 2791, 5932, 6307, 6369 (January to October 2026); config repo issue (October 2026); AINOTE (August 2026); community post (May).

Not run by DIY Lab Hub: we read the files and docs but did not install the plugin on a test machine.

Checked against the official docs on 2026-10-27