Anthropic Adds Plugin Evals to Claude Code: 6 Grader Types, a No-Plugin Baseline, and a CI Gate for Skills

AITopTools Editorial TeamSeptember 11, 2026

What changed

Anthropic has published a plugin-evaluation workflow for Claude Code. The new `claude plugin eval` command tests plugins with realistic prompts, grades Claude’s results, and compares them with results from a run without the plugin. It helps developers measure whether a plugin activates and improves responses.

What this means for you

Plugin developers can use these checks to catch problems before release and add them to continuous integration, which automatically tests changes. The feature is aimed at Claude Code users building or maintaining plugins, rather than everyday users.

Related AI tools

Explore directory listings connected to the products, companies, and workflows in this story.

CodeAcademy Codenomicon codeflying Claude

Related AI news

New AI featuresSep 12, 2026

GPT-6 Astra needs leaner prompts and fewer guardrails, OpenAI recommends

Teams building tools with GPT-6 Astra may be able to simplify their instructions and reduce unnecessary approval steps. The feed does not state the model’s availability, pricing, or access limits.

The DecoderSee why it matters

New AI featuresSep 11, 2026

Can LLMs Engineer Their Own Agent Harness? ByteDance Seed’s HarnessDev Says Only 34 of 64 Changes Generalize

The findings suggest AI models can improve the tools and instructions they use for some tasks, but those improvements may not transfer reliably to unfamiliar work. HarnessDev is a research benchmark, not a new consumer product, so it offers evaluation rather than immediate access to a new capability.

MarkTechPostSee why it matters

New AI featuresSep 11, 2026

Cognition helps Devin test its own work with GPT‑6 Astra

This could make Devin more useful for engineers by giving it a stronger role in checking its own work. The feed does not provide details about availability, pricing, or access.

OpenAI NewsSee why it matters