@kyh/skills@0.1.2

High-risk behavior combination matched malicious policy.

`npm install -g @kyh/skills` runs the postinstall hook.

Package and third-party content becomes available to Codex, Claude, and other agents without a separate setup command or per-project approval.

Lifecycle-driven global AI-agent configuration and extension installation.

On global installation, the postinstall hook links bundled skills, agents, and global instructions into `~/.agents`, mirrors them into `~/.claude`, and automatically installs skills from 14 external repositories using `skills add ... -y`. These locations are global AI-agent control surfaces, so the lifecycle hook causes broad agent behavior/configuration changes without an explicit post-install setup action. The bundled MCP configuration is empty, but the automatic extension installation remains a concrete control-surface mutation.

The package is transparent about its behavior, but the behavior is still an unconsented global-install lifecycle mutation of broad AI-agent control surfaces. Under the firewall policy, that requires blocking even without credential theft or a non-empty MCP server.