Security audit

brain-design-benchmark

Security checks for vulnerabilities and agentic risk

Overview

This is a text-only evaluation framework for AI brain/agent architecture ideas, with no code execution, credential handling, persistence, or destructive behavior found.

Before installing, consider whether you want a broadly routed architecture-evaluation skill active; narrowing its triggers would reduce accidental activation. No evidence was found of code execution, data access, credential use, persistence, or exfiltration.

SkillSpector

Vulnerability Patterns

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

Findings (2)

Vague Triggers

Medium

Confidence: 93% confidence
Finding: The trigger keywords are extremely generic terms like "brain", "design", and "benchmark", which are likely to match many ordinary user conversations unrelated to this skill. In an agent skill-routing system, this can cause unintended activation, context hijacking, or incorrect delegation, increasing the chance that the wrong skill influences responses or captures user input unexpectedly.

Vague Triggers

Medium

Confidence: 83% confidence
Finding: The manifest description embeds a long trigger phrase list containing broad, natural-language expressions such as '我的架构行不行' and general architecture review wording. This can cause unintended invocation by matching ordinary user phrasing, which may route requests into this skill even when the user did not explicitly intend to use it. In this skill’s context, the result is mainly misrouting or over-triggering rather than direct code execution or data compromise.

Static analysis

No suspicious patterns detected.