The automatic postinstall mutation of AI-agent skill directories is a real extension lifecycle risk, but the installer only links package-owned skills and the inspected content is product-aligned. Classify as a warning rather than a confirmed control hijack.
Static reason
High-risk behavior combination matched malicious policy.
Trigger
An npm installation that runs the package's postinstall hook.
Impact
An AI agent that loads skills from those directories may use the package's instructions. The inspected skill content concerns this design system's components and package workflow.
Mechanism
The installer links each packaged skill into the consumer's .claude/skills and .agents/skills directories, replacing an existing symlink at the same destination.
Evidence
package.jsonscripts/install-skill.mjsskills/design-system/SKILL.md.claude/skills.agents/skills