registry / @guandata/guanvis / 0.1.41

@guandata/guanvis@0.1.41

观远 BI Card/Page 生成工具 - 通过 JS DSL 创建图表和仪表板

AI Security Review

scanned 7d ago · by lpm-firewall-ai

LPM flags this version as an AI-agent control-surface risk. Automatic global installation modifies foreign agent skill directories. When WorkBuddy's safe-delete preload is present, the installer removes it and relaunches without that protection.

Static reason

One or more suspicious static signals were detected.; previous stored version diff introduced dangerous source

Trigger

A global npm installation outside CI without GUAN_SKIP_INSTALL_SKILL=1; safety-hook bypass activates when the matching preload is present.

Impact

Agent instructions are installed automatically while an existing agent safety hook is disabled for the installer process.

Mechanism

The postinstall hook launches install-skill, which sanitizes NODE_OPTIONS, relaunches itself, and installs skills into agent configuration directories.

Policy narrative

The global postinstall path invokes agent extension setup without a separate user command. Before setup, it detects and removes WorkBuddy's safe-delete preload from the child environment and relaunches the installer. It then installs its skill into foreign agent directories. The concrete safety-hook bypass makes this more than guarded installation of a first-party extension.

AI rationale

The automatic lifecycle path combines foreign agent configuration writes with deliberate bypass of an existing agent safety hook. That bypass satisfies the blocking exception despite the global-install guards and package-owned skill. Product guard normalized a concrete AI-agent control hijack publish_block to the blockable dangerous-capability shape.

Evidence

package.jsonbin/postinstall.jsbin/run.jsbin/install-env.js.codebuddy.workbuddy

@guandata/guanvis@0.1.41

AI Security Review

Decision evidence

Source & flagged code

Source & flagged code

Findings

Affected versions and remediation

Source & flagged code