What it does

Anthropic open-sourced 11 “plugins” — file-based bundles of markdown instructions, slash commands, and tool connectors — that shape Claude’s behavior for specific job functions. Install one and Claude suddenly knows how your sales team preps for calls, how legal triages NDAs, or how finance runs month-end close. Each plugin wires to external tools (Slack, Salesforce, Snowflake, Jira, etc.) via MCP servers.

The interesting bit

There’s no code. The entire system is markdown and JSON — skills, commands, and connector manifests — which means customizing them is just editing text files. The “plugin” is essentially a very long, very structured prompt that Anthropic version-controls for you.

Key highlights

• 11 role-specific plugins covering sales, legal, finance, data, marketing, product management, customer support, productivity, enterprise search, bio-research, and plugin-management itself
• Skills activate automatically when relevant; slash commands (e.g., /sales:call-prep) are explicit triggers
• Connectors span 30+ tools including CRMs, data warehouses, design tools, and life-sciences databases
• Compatible with both Claude Cowork and Claude Code via CLI install
• Includes a meta-plugin for creating or customizing plugins for roles Anthropic hasn’t covered

Caveats

• The README describes customization as “edit files” but doesn’t specify how those changes propagate — whether they sync back to Cowork, require re-install, or live in a fork is unclear
• Bio-research plugin has the most specialized connectors (PubMed, ChEMBL, Benchling, etc.); value for non-bio teams is obviously zero

Verdict

Worth a look if you’re already in the Claude ecosystem and tired of repeating your company’s processes in every chat. Skip it if you’re not using Claude Cowork or Claude Code — these are prompt templates, not standalone tools, and the 18,964 stars reflect Anthropic’s brand more than universal utility.