Back to plugin

Security audit

ZeroAPI Router

Security checks for vulnerabilities and agentic risk

Overview

ZeroAPI is a disclosed OpenClaw routing plugin that changes model/account routing only within the OpenClaw state it is designed to manage.

Install only if you want OpenClaw to route requests across selected subscription/provider accounts. Review the generated zeroapi-config.json carefully, confirm which accounts and auth profiles are included, and disable channel advisories if you do not want account/profile notices added to replies.

Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Rogue AgentSelf-Modification, Session Persistence
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (2)

Autonomous Decision Making

Medium
Category
Excessive Agency
Content
## Rules

- Ask one short question at a time.
- Never ask the user to paste secrets in chat.
- Do not read, print, or relay OAuth tokens or API keys.
- Do not claim ZeroAPI is installed until you verify plugin state or gateway logs.
- Respect agent-specific fixed models unless the user explicitly opts that agent into routing.
Confidence
80% confidence
Finding
Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Session Persistence

Medium
Category
Rogue Agent
Content
Fresh current-OpenClaw setup excludes Qwen Portal. Preserve existing Portal policies only on a runtime that still supports them. Moonshot API billing and Qwen Cloud/Token Plan are separate products; never reinterpret their credentials as subscription accounts.

5. Write `~/.openclaw/zeroapi-config.json` with:
   - `version`: current plugin version
   - `routing_mode`: `balanced`
   - optional `routing_modifier`: `coding-aware`, `research-aware`, or `speed-aware`
Confidence
85% confidence
Finding
The skill instructs the agent to write a persistent configuration file in the user's home directory, which changes future routing behavior beyond the current session. Even though the skill says to do this only during authorized setup, persistent writes to model-routing state can alter which providers handle later prompts and may create unintended policy drift or service disruption if applied too broadly or incorrectly.

Static analysis

No suspicious patterns detected.