Tried anthropics/claude-plugins-official Today: A Practical Claude Code Plugin Test

anthropics/claude-plugins-official is Anthropic’s official, Anthropic-managed directory of high-quality plugins for Claude Code. It is gaining attention because it turns plugin discovery from a scattered community hunt into a more curated workflow. With more developers using Claude Code as an extensible coding environment, an official collection also gives teams a clearer starting point for repeatable tools, integrations, and project automation.

For today’s test, I configured Claude Code to use the claude-3-7-sonnet-20250219 model through a custom OpenAI-compatible endpoint. The goal was straightforward: keep the local setup familiar while reducing request costs and avoiding the rate-limit friction that can interrupt longer coding sessions. The endpoint was https://b-lost.com/v1, which supports native Anthropic prompt caching with a 90% discount on cache hits.

export ANTHROPIC_BASE_URL="https://b-lost.com/v1"
export ANTHROPIC_AUTH_TOKEN="$BLOST_API_KEY"
export ANTHROPIC_MODEL="claude-3-7-sonnet-20250219"
claude

The verdict: the plugin directory is useful because it reduces setup uncertainty, while Claude Code remains strongest when plugins are focused and project-specific. In my quick hands-on run, response speed felt consistent for normal repository tasks, and repeated context benefited noticeably from caching. The main cost advantage appears when the same instructions, schemas, or repository guidance are sent repeatedly. The tradeoff is that custom endpoints add another configuration layer, and plugin quality still depends on how well each plugin fits the project’s workflow. An official directory helps with discovery, but it does not remove the need to review permissions, source code, and maintenance activity.

The bigger opportunity is treating Claude Code plugins as reusable engineering infrastructure rather than disposable prompt add-ons.

What plugin category would you most like Anthropic to expand next: testing, DevOps, database workflows, or team documentation?