registry / agent-tower / 0.5.3

agent-tower@0.5.3

AI Agent Task Management Dashboard

AI Security Review

scanned 3mo ago · by lpm-firewall-ai

No confirmed malicious attack surface. The package is an AI agent task dashboard with expected runtime capabilities: local server, SQLite state, PTY-backed agent sessions, MCP API proxy, git/worktree operations, and optional user-started tunnel.

Static reason

High-risk behavior combination matched malicious policy.

Trigger

Install runs postinstall; runtime behavior requires invoking agent-tower, agent-tower-mcp, or dashboard/API actions.

Impact

Can spawn configured agent CLIs and expose the local dashboard through a token-protected tunnel when the user starts it; no unconsented exfiltration or control-surface mutation was found.

Mechanism

package-aligned local dashboard and agent orchestration

AI rationale

Source inspection shows high-risk primitives are part of the advertised agent dashboard and are user-invoked, while the install hook is limited to fixing node-pty helper permissions. I found no concrete malware behavior, credential/file harvesting, hidden network exfiltration, or unconsented AI-agent control mutation.

Evidence

package.jsonscripts/postinstall.jsdist/cli.jsdist/index.jsdist/mcp/index.jsdist/mcp/server.jsdist/services/tunnel.service.jsdist/services/session-manager.jsdist/services/terminal-manager.jsnode_modules/**/node-pty/**/spawn-helper~/.agent-tower/data.db~/.agent-tower/port~/.agent-tower/*.log

agent-tower@0.5.3

AI Security Review

Decision evidence

Source & flagged code

Source & flagged code

Findings

Affected versions and remediation

Source & flagged code