Browse the full directory of Claude Code plugins — commands, agents, skills, MCP servers, and more.
Browse plugins →Search “caveman skill” and Claude Code returns the same plugin eight times. Here's why they're identical and how to pick which copy to install.
Search Google for caveman skill and you will find the caveman plugin listed under a string of different owners, all tracing back to the same caveman-engineering source. It shows up in this directory as eight separate listings, each shipping the same three component types — commands, agents, and skills — with nothing in the data to tell them apart. You are not choosing between rival caveman skills; you are choosing which collection to install the same one from.
The verdict is short:
flight505 copy if you want the listing that labels itself upstream — the source the others were forked from.ai-integr8tor's copy if you want caveman bundled inside the widest sibling collection, a repo of 78 plugins.richyboy170's copy if you want the leanest surroundings; its repo carries 35.The query is not niche: "caveman skill" drew 15,881 impressions and 66 clicks over 28 days at an average Google position of 8.4, which is why it is worth being precise about what the directory actually returns under that name.
Every one of the eight carries the same trio: commands, agents, and skills. Those are three of the Claude Code component types — the same building blocks you would assemble to automate a Claude Code workflow end to end. None of the eight adds hooks, an MCP server, or anything else that would set one copy apart.
What is missing is any way to rank them. None has a plugin-level star count. Each lives inside a large multi-plugin repository, and a repository's GitHub stars belong to the whole collection, not to the caveman plugin buried inside it — so a star number here would be borrowed credit, not a signal. There are no per-plugin install counts in the snapshot either. When you open any of these eight pages, that component list is essentially the whole spec the directory can show: no maintainer-written summary, no freshness or last-updated signal to prefer one copy over another. The usual tiebreakers are gone, and the only field that varies across the eight is the size of the repo each copy ships inside.
| Listing (owner) | Source repo hosts | Components |
|---|---|---|
| ai-integr8tor | 78 plugins | commands · agents · skills |
| kruxshnx | 65 plugins | commands · agents · skills |
| msm47 | 63 plugins | commands · agents · skills |
| marco3939 | 60 plugins | commands · agents · skills |
| alexbramall | 58 plugins | commands · agents · skills |
| ciciliaeth | 43 plugins | commands · agents · skills |
| flight505 (upstream) | 43 plugins | commands · agents · skills |
| richyboy170 | 35 plugins | commands · agents · skills |
All eight copies come from repositories named caveman-engineering, each of which hosts dozens of other plugins — 78 in ai-integr8tor's, 65 in kruxshnx's, 35 in richyboy170's, and counts in between for the rest. These are aggregation repos: an owner collects a batch of Claude Code plugins and skills into one marketplace, and caveman is one of the entries that keeps getting re-listed. That is why a single searchable skill fans out into eight near-identical directory pages, and it also explains the missing metrics — the stars on each source repo were earned by the whole collection, and the directory correctly declines to credit them to caveman alone.
Because the plugins are identical, the honest way to separate them is by the repository around each one: installing a copy means pulling it from that owner's marketplace, where its sibling plugins live too. Largest repo first.
The widest surrounding collection of the eight, so it carries the most other tooling alongside the same caveman skill. Reach for this listing if that owner's other plugins are ones you would also install.
Second-largest surroundings, identical caveman entry. Install this copy if kruxshnx's collection is already part of your setup.
No difference in the caveman plugin itself — only the 63-plugin repo it rides in. Use this listing if you want caveman from that particular collection.
Middle of the pack on repo size, same commands-agents-skills skill. Install this copy if you are already pulling from marco3939's collection.
Another identical copy, this time from a 58-plugin repo. The listing is here if that marketplace fits your stack.
One of two copies at the smaller end of the range. Take this one if ciciliaeth's other plugins are ones you would use.
The only listing whose slug marks it upstream, which makes it the natural pick if you want the source the others were forked from — even though its repo, at 43 plugins, is one of the smaller ones. Install it here.
The smallest surrounding collection of the eight, so the least extra baggage if all you want is caveman and nothing else. The listing is here.
Since the eight are the same skill, the decision is really about which marketplace you want to live in. If you already install from one of these owners' collections, take the caveman copy that ships with it. If you don't, take the flight505 upstream copy. One caution: because each copy is published from a broad, multi-purpose repo rather than a dedicated caveman project, treat the surrounding collection — not the caveman entry — as what you are really evaluating; if an owner's other plugins look off-topic for your work, prefer the upstream copy and skip the aggregation repos. And if what you actually need is to know how caveman behaves before committing, these eight listings won't tell you — the snapshot carries no description for any of them, and the caveman page that most "caveman skill" clicks land on is a separate listing entirely. Open /plugins, search caveman, and compare these copies against that one before you install.
Eight Claude Code plugins for Godot compared by component mix: GDScript skill packs, MCP editor control, hooks, and multi-engine game-dev bundles.
A data-anchored walk through the development category on ClaudePluginHub — 6,299 plugins — covering frontend auditing, full-stack skills, model guardrails, prompt management, and CMS tooling, with install and star counts for each.
Walk through eight plugins listed in the deployment category, with 7-day install counts, GitHub stars, and component coverage — from UI/UX skills to documentation lookup, codebase graphs, and SEO automation.
Ultra-compressed communication mode. Cuts token usage ~75% by dropping filler, articles, and pleasantries while keeping full technical accuracy. Enhanced from Matt Pocock's MIT-licensed caveman skill (https://github.com/mattpocock/skills) with: (1) stdlib Python tools (text compressor, token-savings estimator, caveman-style linter), (2) 3 reference docs citing 5+ authoritative sources each (compression principles, technical communication patterns, when caveman backfires), (3) cs-caveman-mode persona agent + /cs:caveman slash command. Matt's voice and persistence rules preserved verbatim per MIT. Use when user says "caveman mode", "talk like caveman", "use caveman", "less tokens", "be brief", or invokes /caveman.
Ultra-compressed communication mode. Cuts token usage ~75% by dropping filler, articles, and pleasantries while keeping full technical accuracy. Enhanced from Matt Pocock's MIT-licensed caveman skill (https://github.com/mattpocock/skills) with: (1) stdlib Python tools (text compressor, token-savings estimator, caveman-style linter), (2) 3 reference docs citing 5+ authoritative sources each (compression principles, technical communication patterns, when caveman backfires), (3) cs-caveman-mode persona agent + /cs:caveman slash command. Matt's voice and persistence rules preserved verbatim per MIT. Use when user says "caveman mode", "talk like caveman", "use caveman", "less tokens", "be brief", or invokes /caveman.
Ultra-compressed communication mode. Cuts token usage ~75% by dropping filler, articles, and pleasantries while keeping full technical accuracy. Enhanced from Matt Pocock's MIT-licensed caveman skill (https://github.com/mattpocock/skills) with: (1) stdlib Python tools (text compressor, token-savings estimator, caveman-style linter), (2) 3 reference docs citing 5+ authoritative sources each (compression principles, technical communication patterns, when caveman backfires), (3) cs-caveman-mode persona agent + /cs:caveman slash command. Matt's voice and persistence rules preserved verbatim per MIT. Use when user says "caveman mode", "talk like caveman", "use caveman", "less tokens", "be brief", or invokes /caveman.
Ultra-compressed communication mode. Cuts token usage ~75% by dropping filler, articles, and pleasantries while keeping full technical accuracy. Enhanced from Matt Pocock's MIT-licensed caveman skill (https://github.com/mattpocock/skills) with: (1) stdlib Python tools (text compressor, token-savings estimator, caveman-style linter), (2) 3 reference docs citing 5+ authoritative sources each (compression principles, technical communication patterns, when caveman backfires), (3) cs-caveman-mode persona agent + /cs:caveman slash command. Matt's voice and persistence rules preserved verbatim per MIT. Use when user says "caveman mode", "talk like caveman", "use caveman", "less tokens", "be brief", or invokes /caveman.
Ultra-compressed communication mode. Cuts token usage ~75% by dropping filler, articles, and pleasantries while keeping full technical accuracy. Enhanced from Matt Pocock's MIT-licensed caveman skill (https://github.com/mattpocock/skills) with: (1) stdlib Python tools (text compressor, token-savings estimator, caveman-style linter), (2) 3 reference docs citing 5+ authoritative sources each (compression principles, technical communication patterns, when caveman backfires), (3) cs-caveman-mode persona agent + /cs:caveman slash command. Matt's voice and persistence rules preserved verbatim per MIT. Use when user says "caveman mode", "talk like caveman", "use caveman", "less tokens", "be brief", or invokes /caveman.
Ultra-compressed communication mode. Cuts token usage ~75% by dropping filler, articles, and pleasantries while keeping full technical accuracy. Enhanced from Matt Pocock's MIT-licensed caveman skill (https://github.com/mattpocock/skills) with: (1) stdlib Python tools (text compressor, token-savings estimator, caveman-style linter), (2) 3 reference docs citing 5+ authoritative sources each (compression principles, technical communication patterns, when caveman backfires), (3) cs-caveman-mode persona agent + /cs:caveman slash command. Matt's voice and persistence rules preserved verbatim per MIT. Use when user says "caveman mode", "talk like caveman", "use caveman", "less tokens", "be brief", or invokes /caveman.
Ultra-compressed communication mode. Cuts token usage ~75% by dropping filler, articles, and pleasantries while keeping full technical accuracy. Enhanced from Matt Pocock's MIT-licensed caveman skill (https://github.com/mattpocock/skills) with: (1) stdlib Python tools (text compressor, token-savings estimator, caveman-style linter), (2) 3 reference docs citing 5+ authoritative sources each (compression principles, technical communication patterns, when caveman backfires), (3) cs-caveman-mode persona agent + /cs:caveman slash command. Matt's voice and persistence rules preserved verbatim per MIT. Use when user says "caveman mode", "talk like caveman", "use caveman", "less tokens", "be brief", or invokes /caveman.
Ultra-compressed communication mode. Cuts token usage ~75% by dropping filler, articles, and pleasantries while keeping full technical accuracy. Enhanced from Matt Pocock's MIT-licensed caveman skill (https://github.com/mattpocock/skills) with: (1) stdlib Python tools (text compressor, token-savings estimator, caveman-style linter), (2) 3 reference docs citing 5+ authoritative sources each (compression principles, technical communication patterns, when caveman backfires), (3) cs-caveman-mode persona agent + /cs:caveman slash command. Matt's voice and persistence rules preserved verbatim per MIT. Use when user says "caveman mode", "talk like caveman", "use caveman", "less tokens", "be brief", or invokes /caveman.