registry / @volcengine/ark-cli / 1.0.20

@volcengine/ark-cli@1.0.20

火山方舟 ARK 平台命令行工具

AI Security Review

scanned 2h ago · by lpm-firewall-ai

LPM flags this version as an AI-agent control-surface risk. npm postinstall downloads a native binary, then automatically runs its agent-skill refresh command. This mutates local AI-agent capability surfaces without an explicit user command.

Static reason

One or more suspicious static signals were detected.

Trigger

npm installation outside CI

Impact

Downloaded code can install Ark Skills into detected local agents during installation.

Mechanism

remote native binary bootstrap followed by automatic agent-skill installation

Policy narrative

On npm install, the postinstall script downloads a platform binary from manifest-controlled CDN/GitHub URLs, verifies its SHA-256, executes it, and then invokes `+connect --refresh`. The package documentation states that `+connect` installs Ark Skills into local agents. This is an unconsented install-time modification of a broad local AI-agent control surface.

Rationale

The checksum check constrains the configured payload but does not remove the unconsented postinstall agent-extension action. The lifecycle hook executes the action automatically rather than requiring the documented user command.

Evidence

package.jsonscripts/postinstall.jsmanifest.jsonREADME.mdbin/skills/

Network endpoints2

lf3-static.bytednsdoc.comgithub.com

@volcengine/ark-cli@1.0.20

AI Security Review

Decision evidence

Source & flagged code

Source & flagged code

Findings

Source & flagged code