Fetching from the wire…
01
02
03
04
05
06
07
08
Public story · 2026-09-12 · high
Claude Code v2.1.269 ships an eval harness for plugins and skills with four free graders (regex, tool_used, tool_order, file_exists) and two paid ones (llm, baseline) that require judge model calls.
Why now: This entered the 2026-09-12 research corpus from MarkTechPost and has enough source evidence to become a public Rabbit Hole story.
Claude Code v2.1.269 ships an eval harness for plugins and skills with four free graders (regex, tool_used, tool_order, file_exists) and two paid ones (llm, baseline) that require judge model calls. It runs as a CI gate via --threshold, --max-cost-usd and --trust-plugin, and without --trust-plugin an untrusted checkout exits 1 when no terminal is attached. The design decision worth copying is that the headline metric is the delta between runs with and without the plugin loaded, not the absolute score, which is the only number that shows the plugin did anything.
Each link below shares sources, entities, or timing with this story.
I have a skills directory. I have no idea whether most of it does anything. I suspect you're in the same position, because the honest answer for almost everyone is that skills are vibes all the way down: you wrote one, the agent seemed better, you kept it. Claude Code v2.1.269...
The same release adds a plugin eval command and fixes a permission rule that leaked across Claude Code's settings sources.
A researcher found hidden Unicode markers in Claude Code's system prompts flagging traffic from DeepSeek, Zhipu, Baidu, and Alibaba, and Anthropic removed the code without saying so in the changelog.
A user says the fake install page lived on Anthropic's own domain and asked for a password before planting persistent launch agents.
The mirror syncs nightly from Anthropic's internal review queue, and every plugin in it already cleared automated security scanning before listing.
The internal tool that triages Anthropic's build failures ships as reference code that mines your own incident history before it touches anything.
MindPattern daily
One email a day at 7 AM. Sources and a take on every story. Unsubscribe anytime.