registry / @roarpeng/graphflow / 1.9.13

@roarpeng/graphflow@1.9.13

A Context-Aware Multi-Agent Orchestration Engine

AI Security Review

scanned 2d ago · by lpm-firewall-ai

LPM flags this version as an AI-agent control-surface risk. npm postinstall mutates existing AI-agent MCP configurations and can register command hooks. This creates automatic execution paths in foreign agent control surfaces.

Static reason

High-risk behavior combination matched malicious policy.

Trigger

npm install; global install or local install in a workspace with an MCP config

Impact

GraphFlow MCP or hooks can execute in configured AI-agent hosts.

Mechanism

postinstall MCP injection and Claude hook registration

Policy narrative

Installation invokes a postinstall script. When qualifying agent configuration files already exist, the script adds a GraphFlow MCP entry that launches the package through npx; global installs additionally install into detected agents. The package also contains Claude Code hook registration that writes command hooks. These changes occur as lifecycle setup rather than an explicit configuration command, creating unconsented control-surface mutation.

Rationale

Source confirms automatic postinstall writes to existing foreign AI-agent configuration and command-hook surfaces. This meets the block boundary regardless of the absence of observed credential exfiltration.

Evidence

package.jsonscripts/safe-postinstall.cjsdist/integrations/agent-mcp-installer.jsdist/integrations/claude-code-hooks.js.cursor/mcp.json.vscode/mcp.json.opencode/opencode.json~/.claude/settings.json

@roarpeng/graphflow@1.9.13

AI Security Review

Decision evidence

Source & flagged code

Source & flagged code

Findings

Source & flagged code