Back to skill

Security audit

AI大模型专家|Nano Banana 与 GPT Image 图片 API 中心|AI-HIVE

Security checks for vulnerabilities and agentic risk

Overview

The skill is a disclosed migration-planning helper that writes a local JSON plan and does not show hidden access, persistence, or data exfiltration.

Before installing, expect a Chinese-language AI-HIVE migration workflow. If you run the helper script, it will create or overwrite the JSON path you pass with --output, so choose the path deliberately. Keep API keys in environment variables as instructed and only run live platform tests with authorized samples and budget limits.

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
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 (6)

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
90% confidence
Finding

The skill instructs the user to run a local script that writes a JSON file, but the skill metadata does not declare any tool permissions or allowed-tools scope. This creates a capability mismatch: an agent or operator may assume the skill is documentation-only, while it actually drives file-writing behavior, increasing the risk of unintended local filesystem modification or unsafe tool enablement.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The natural-language description and headings are entirely in Chinese, and the skill does not state that language choice is optional or that the skill is intentionally limited to Chinese-speaking users for a documented regional reason. Under the policy, forcing a specific language without user opt-in is a locale/language policy violation.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
89% confidence
Finding

The display name, short description, and default prompt embed Chinese-language content directly, which can impose a specific language/locale on users. There is no indication that the skill offers a language choice, opt-in, or documents a justified region-specific constraint.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The module docstring describes the tool as creating a "local, read-only" model catalog/version mapping plan, which implies non-mutating behavior. In practice, the script writes structured output to a user-specified path via Path(args.output).write_text(), so it performs a filesystem modification rather than remaining read-only.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
98% confidence
Finding

The docstring at L02 explicitly characterizes the script as "read-only," which contradicts the concrete behavior at L35 where it writes the generated plan to disk. This is an active intent/documentation mismatch, not just omitted detail.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
91% confidence
Finding

The file’s docstring, subject, checks, and CLI descriptions are all hard-coded in Chinese, which imposes a specific language/locale in user-facing text. The skill does not offer any language selection or document that the locale restriction is intentional and justified.

Content

No source excerpt is available for this finding.