Security audit

Enterprise AI Governance Playbook（企业AI治理实操手册）

Security checks for vulnerabilities and agentic risk

Overview

This skill is a disclosed AI governance playbook with a user-run local Python helper and no evidence of hidden network, credential, persistence, or destructive behavior.

Installers should understand that the skill includes user-run Python utilities. Use it for governance drafting and self-assessment, keep sensitive company details out of prompts or command arguments unless your environment is appropriate, and verify legal/regulatory summaries with qualified counsel before relying on them.

SkillSpector

Vulnerability Patterns

MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
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)

Lp3

Medium

Category: MCP Least Privilege
Confidence: 93% confidence
Finding: The skill advertises executable local tooling that reads and writes files and is invoked via shell commands, yet no explicit permissions are declared. This creates a trust and enforcement gap: users or hosting platforms may not realize the skill can access the filesystem and execute commands, reducing oversight and increasing the chance of unsafe use or privilege overreach.

Intent-Code Divergence

Medium

Confidence: 97% confidence
Finding: The docstring claims the scan does not record real identifiers, but the implementation stores part of each match via m.group(0)[:20] in the hits list. If secrets, emails, tokens, or private-key material are detected, even truncated values may expose sensitive data in memory, logs, console output, or downstream JSON/reporting paths.

Static analysis

No suspicious patterns detected.