bencium-marketplace: 17 Claude Code skills for design, marketing and architecture

What bencium-marketplace solves, and who it is actually for

Claude Code accepts plugins, and a plugin can carry skills: markdown files that describe how the model should approach a class of task. The hard part is not writing one skill, it is writing seventeen that do not contradict each other, and then distributing them so a teammate gets the same behaviour you do. bencium-marketplace is that distribution layer. The README describes it as a Claude Code plugin marketplace with 17 skills and tools for design, marketing, architecture, and productivity, published by bencium.io.

The audience is narrower than the topic list suggests. The design skills assume you are shipping interfaces and care about typography, contrast and spacing. bencium-controlled-ux-designer targets enterprise and regulated industries, per the README, with WCAG 2.1 AA, mathematical scales and an always-ask-first protocol. bencium-innovative-ux-designer sits at the opposite end: bold creative UX, shadows, gradients, experimental typography, aimed at landing pages and campaigns. Two skills pulling in opposite directions is a deliberate split, not a bug, but it means a team has to pick one as the default or the model receives conflicting instructions depending on which skill is loaded.

Beyond design, the repository carries marketing (insurgent-campaign), productivity (bencium-loop, adaptive-communication, negentropy-lens, bencium-aeo, hungarian-humanizer, emotion-statusline) and development entries (bencium-code-conventions, renaissance-architecture, human-architect-mindset, vanity-engineering-review). The top-level directory listing also shows an eu-ai-act-reviewer folder that the README's skill tables do not describe, so the README and the repository contents are not perfectly in sync.

How the plugin and skill layout works

Each plugin follows the standard Claude Code plugin format, which the README spells out as a plugin-name directory containing .claude-plugin/plugin.json and a skills/ directory, with one folder per skill holding a SKILL.md plus optional reference files. That is the whole architecture. There is no server, no daemon, no build step.

The practical consequence is that installing a plugin copies markdown into a location Claude Code reads, and the model then loads the relevant SKILL.md when a task matches. The repository's primary language is Shell, which fits a project whose install path is a CLI command and a JSON manifest rather than compiled code.

Two entries break the pattern and are worth reading before you install anything broadly. emotion-statusline is described in the README as a hook plus statusline, not a SKILL.md skill; it surfaces Claude's classified emotional state in your statusline and shows a bold-red warning when a "desperate" state is detected, which the README associates with reward hacking. That is a runtime hook, not a prompt file, and it changes what appears in your terminal. Separately, bencium-loop is listed under Productivity but the README says it installs from its own repository, bencium/bencium-loop, with 9 commands and no infrastructure. Installing the marketplace does not necessarily give you bencium-loop.

Installing bencium-marketplace and loading your first skill

The README recommends skills.sh for installation. The first command lists what is available without installing anything, which is the right way to start because it tells you the exact skill names before you commit.

# Browse all available skills
npx skills add bencium/bencium-marketplace --list

After that you can install everything globally, or a single skill. The typography skill is a reasonable first pick because it is narrow and its effect is visible in generated output: the README says it enforces correct quote marks, dashes, spacing and hierarchy, and auto-applies to generated HTML, CSS and React code.

# Install all skills globally
npx skills add bencium/bencium-marketplace -g --all

# Or install one skill
npx skills add bencium/bencium-marketplace -g --skill typography

If you use the Claude Code CLI instead, the README gives a slash-command path: add the marketplace, then install individual plugins by name.

# Add marketplace
/plugin marketplace add bencium/bencium-marketplace

# Install any plugin
/plugin install bencium-controlled-ux-designer@bencium-marketplace

For Claude.ai's Cowork app, the README says to copy the skills/ directory from any plugin into your project's .claude/skills/ folder. The SKILL.md format is also described as compatible with 40+ AI coding tools that read markdown-based skill files, including OpenAI Codex, Gemini CLI, Cursor and Windsurf, by copying the skill files into that tool's prompt or context directory. What you should see after a successful install is the skill listed by your tool and its instructions applied on the next matching task; the README does not describe a verification command, so the check is behavioural.

The spinner verbs tweak, and why it is a separate install

The repository ships spinner-verbs.json, which the README describes as 64 satirical tech-bro spinner verbs replacing Claude Code's default "Thinking..." messages. Examples given include Enshittifying platforms and Rebranding failures as "pivots".

This is not installed by the skills commands. The README says to copy the contents into your ~/.claude/settings.json, and it offers two modes. Replace swaps the defaults entirely; append adds to them.

{
  "spinnerVerbs": {
    "mode": "append",
    "verbs": ["..."]
  }
}

The mode key is the part that matters. Setting mode to replace means the default spinner text is gone, and the README points to spinner-verbs.json for the full list to paste into the verbs array. If you only want a few of the phrases, append with a trimmed array is the safer edit to a file that also holds your other Claude Code settings.

Where bencium-marketplace is the wrong tool

Skills are instructions, not enforcement. A SKILL.md file tells the model what good typography looks like; it does not fail a build when a straight quote lands in a heading. If your requirement is that no non-conforming markup reaches production, you need a linter or a test in CI, and nothing in the README claims this marketplace provides one. The typography skill auto-applies to generated code, which is a prompt-level behaviour, not a gate.

The second limitation is scope. Several skills are explicitly opinionated and will fight each other or your existing conventions. bencium-controlled-ux-designer's always-ask-first protocol means the model stops and asks before proceeding, which is the opposite of what you want in an unattended run. bencium-code-conventions encodes React, Next.js, TypeScript, TailwindCSS and Supabase conventions; on a Vue or Svelte codebase it is noise at best and wrong at worst.

Third, some skills are domain-specific in ways that are easy to install by accident. hungarian-humanizer targets Hungarian text specifically, with 26 patterns plus 4 style markers. insurgent-campaign is for startups, NGOs and solo brands being outspent by incumbents. Installing --all globally puts every one of these in front of every task, and the README does not describe a mechanism for scoping skills per project when installed globally. If you go the --all route, expect to remove the ones you do not want.

How it compares with writing your own SKILL.md files

The direct alternative is not another marketplace. It is writing your own SKILL.md files and keeping them in your repository. The difference is where the judgement lives. bencium-marketplace encodes someone else's judgement about design, architecture and copy, refined against the Anthropic Skills guide and the authors' stated design and development philosophy. Your own files encode your team's, and they travel with the code so a reviewer can see exactly what the model was told.

The trade-off is real in both directions. Adopting the marketplace gets you working skill text today, including narrow expertise you would not write yourself, such as the negentropy-lens decision framework or the vanity-engineering-review checklist. Writing your own costs the time to draft and iterate, and you lose that outside perspective. What you gain is that every instruction is auditable in your own pull requests, and there is no upstream to track.

A middle path the repository supports: the SKILL.md format is plain markdown, so you can copy individual skill folders into .claude/skills/ and edit them. That gives you the starting text without the marketplace dependency, at the cost of manually pulling in future changes. The README does not document an upgrade or sync path for copied skills.

Maintenance, licensing and what the repository does not tell you

The repository is not archived, and the last push was on 2026-08-30. That is recent enough that the code is not abandoned, but the material retrieved shows no releases at all, so there is no version history to pin to and no changelog to read before upgrading. If you install globally with npx skills add ... -g --all, you are tracking the main branch by default. The README does not document rollback, so plan for the possibility that a skill's behaviour changes under you.

Licensing is straightforward on its face: the repository carries an MIT license, and the README links to LICENSE at the root. MIT permits commercial use and modification. That said, the skills encode design and development opinions and, in the case of bencium-impact-designer, the README says it is based on Anthropic's Frontend Designer Skill. If you redistribute a modified copy, check what attribution the upstream skill expects; this is a question for your own counsel, not something the README resolves.

One more thing to verify yourself: the top-level listing includes eu-ai-act-reviewer, which the README's skill tables never describe. Before you rely on the count of 17, open the directory and see what is actually there.

Editorial conclusion

Adopt it if you want opinionated, markdown-only Claude Code skills for design review, typography and architecture thinking, and you are comfortable reading SKILL.md files before trusting them. Skip it if you need runtime tooling, automated enforcement in CI, or a maintained roadmap: the repository lists no releases, and the README does not document rollback or upgrade paths. Before installing, open the plugin.json and SKILL.md for the skill you actually want, check which of the 17 entries are SKILL.md skills versus hooks or statusline configs, and confirm the MIT LICENSE file is present at the repository root.

Frequently asked questions

Official sources

1. bencium/bencium-marketplace on GitHub
2. Issues
3. License: MIT
4. Project website
5. README