Back to skill

Security audit

ai-capabilities

Security checks for vulnerabilities and agentic risk

Overview

This skill is a disclosed AI capability wrapper whose file writes, local persona reads, and API key use fit its stated persona, voice, and image-generation purpose.

Install only if you want a broad AI-capability router that can call configured MCP services, use your SiliconFlow key for generation, and update local persona state files such as SOUL.md. Review persona-changing requests before running them because they can affect future agent behavior.

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 (2)

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The skill metadata includes a very broad trigger list such as 'AI能力', '语音合成', and '图像生成' without clear scope boundaries or disambiguation rules. This can cause over-activation in unrelated conversations, increasing the chance that a high-privilege skill with read/write and AI-generation capabilities is invoked unexpectedly and performs actions the user did not intend.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The skill declares write capability and reliance on sensitive configuration/environment data, including file-modifying tools and SILICONFLOW_API_KEY, but the user-facing documentation does not warn that the skill may modify local persona/state files or consume privileged external-service credentials. In this context, the skill also exposes tools that explicitly write to SOUL.md and likely other state files, making silent invocation or indirect use more risky.

Content

No source excerpt is available for this finding.