The automatic postinstall changes user-level AI agent skill directories during qualifying global installs, which leaves a first-party extension lifecycle risk. The inspected payload is a product skill, so the evidence supports a warning rather than a malicious block.
Static reason
High-risk behavior combination matched malicious policy.
Trigger
Installing globally with npm or pnpm, or using the supported pnpm global-add context.
Impact
Detected agents may load and follow the package's CLI instructions without a separate skill installation.
Mechanism
The postinstall hook calls the skill synchronizer, which copies the bundled skill into the canonical skill directory and directories for detected AI hosts.
Evidence
package.jsondist/bin/postinstall.jsdist/src/features/skill/installer.jsskills/gd-cli/SKILL.md