Back to skill

Security audit

Model Throughput Tester

Security checks for vulnerabilities and agentic risk

Overview

This skill is a coherent model-speed benchmarking tool whose network calls, local report writing, and OpenClaw CLI use match its stated purpose.

Installers should understand that running this skill may send benchmark prompts to OpenClaw or to an API endpoint they provide, and reports may be saved locally. Use non-sensitive test prompts, trusted API URLs, and avoid putting long-lived secrets directly on command lines when possible.

Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (4)

Missing User Warnings

Medium
Confidence
93% confidence
Finding
The README instructs users to send prompts to external API endpoints and generate local reports, but it does not warn that benchmark prompts, model outputs, and possibly sensitive data may be transmitted to third-party services and written to disk. In a benchmarking skill, this omission can cause unintentional data exposure because users may test with real prompts or use production endpoints without understanding the privacy and persistence implications.

Missing User Warnings

Medium
Confidence
88% confidence
Finding
The README instructs users to call an external OpenAI-compatible API with a URL and API key, but it does not clearly warn that prompts, outputs, and credentials may be sent to a third-party service outside the local session context. In an agent skill context, this omission is security-relevant because users may assume the tool is a harmless local benchmark, when in fact it can transmit sensitive prompts or use privileged credentials against arbitrary endpoints.

Vague Triggers

Medium
Confidence
93% confidence
Finding
The manifest includes broad English trigger terms such as "benchmark" and generic speed-testing phrases in a skill that can execute benchmarking commands. These terms are likely to match ordinary user conversations about model evaluation or performance, causing unintended skill activation and potentially unexpected command execution or network/API use.

Vague Triggers

Medium
Confidence
96% confidence
Finding
The trigger list contains highly ambiguous single-term activations, especially "tokens/s", "测速", and "benchmark", which are common in normal discussion and not reliably indicative of intent to run this skill. In context, this is more dangerous because the skill can invoke local Python scripts and, in API mode, send requests to external endpoints with user-supplied keys or URLs if accidentally routed.

Static analysis

No suspicious patterns detected.