Claude Code Marketplace vs SkillzBench: Which Agent Skills Platform Should You Trust?

The Claude Code Marketplace is Anthropic's official mechanism for discovering and installing plugins for Claude Code, its terminal-based agentic coding tool. Anthropic describes plugins as bundles that can include skills, agents, hooks, MCP servers and other capabilities, and the official docs say the marketplace lets you discover and install those extensions without building them yourself. The official marketplace, claude-plugins-official, is automatically available when you start Claude Code, and installation is a single command (/plugin install) or a browse through the Discover tab. Third-party marketplaces such as Awesome Claude Plugins can be added with /plugin marketplace add. The ecosystem has grown large: community directories track dozens of collections spanning LSP code-intelligence plugins, MCP integrations for GitHub, Linear, Slack, Vercel and Sentry, code-review plugins, and security plugins such as Security Guidance and Semgrep. The marketplace is specific to Claude Code, which itself requires a paid Anthropic plan and is terminal-first.

pricing See pricing

• + Live and built in: the official marketplace is automatically available when you start Claude Code, with no separate signup.
• + Very fast installation — a single slash command or a browse through the Discover tab.
• + Official, Anthropic-managed directory described as high quality, plus a large third-party ecosystem (community collections list hundreds of skills, agents and commands).
• + Plugins package reusable workflows — brand guidelines, docs lookup, code review, security guardrails — instead of rebuilding custom commands and MCP configs each time.
• + Strong plugin variety: LSP code intelligence, MCP integrations (GitHub, Linear, Slack, Vercel, Sentry, Supabase), code review and security plugins.

• − Tied to Claude Code, which is terminal-first, Anthropic-models-only, and has no free tier — sustained agent use pushes users toward expensive Max tiers, per third-party reporting.
• − No independent security vetting or benchmarking of third-party skills and plugins is described; users largely trust the packager.
• − Installing many MCP integration plugins can choke the LLM context window by loading all available tools, a caveat flagged by plugin guides.
• − Quality varies across the dozens of third-party marketplaces and collections; there is no unified trust score or measured performance data.
• − Ecosystem scale makes discovery hard: community directories track 34+ separate collections with overlapping contents.

https://github.com/anthropics/claude-plugins-official

SkillzBench is AIQ Labs' free, independent directory of AI agent skills, plus a free MCP tool that connects any agent to the skills you choose — currently in active development ahead of launch, with the waitlist and newsletter live. It is being built around five pillars: Discover (continuously crawling and deduplicating skills from GitHub), Verify (pinning every version to a commit SHA, then validating, license-checking, security-scanning and quality-judging each skill), Benchmark (testing every skill in a sandbox with the same model, with and without the skill, and publishing the Lift in percentage points), Equip (one free MCP server plus per-agent loadouts, where users toggle skills and tools per agent and the agent only sees what was chosen), and Stay current (tracking upstream changes, re-scanning, re-benchmarking and notifying users). SkillzBench will publish a 0-100 trust score built from security, quality, license, maintenance and community signals, and it will keep Lift shown separately from ratings so money never touches scores. The motivation is concrete: published research found roughly 26% of skills carry at least one vulnerability. Everything on the platform — the directory, the MCP connector, benchmarks and alerts — will be free, with revenue coming from advertising around blog articles.

pricing 0

• + Agent-agnostic by design: skills are SKILL.md instruction packages usable with Claude Code, Codex, Cursor, Gemini CLI and chat-app agents.
• + Trust-first approach: commit-SHA version pinning, license checks, security scans and a 0-100 trust score directly address the finding that ~26% of skills carry at least one vulnerability.
• + Measured performance: sandbox benchmarking with the same model, with and without each skill, publishing Lift in percentage points — data the marketplace ecosystem doesn't provide.
• + Everything free — the directory, MCP connector, benchmarks and alerts — with advertising revenue that never touches scores or listings.
• + Per-agent loadouts mean an agent only sees the skills and tools you explicitly toggle on, reducing context bloat and exposure.
• + Ongoing monitoring: upstream changes are tracked, skills are re-scanned and re-benchmarked, and users are notified.

• − Pre-launch: the product is not yet public, so none of its benchmarks, trust scores or the MCP connector can be used today.
• − As a new, independent directory, it will not have Anthropic's first-party integration into Claude Code's own installation flow.
• − Benchmark Lift depends on sandbox testing with a single model, which may not perfectly predict results in every business's own environment.
• − Advertising-funded model means users will see ads in and around blog content (though never tied to scores).

https://skillzbench.com