Back to skill

Security audit

AI大模型专家|AI API 高可用回退路由|AI-HIVE

Security checks for vulnerabilities and agentic risk

Overview

This skill is a disclosed AI API failover planning aid with a small local JSON template generator, not evidence of hidden or destructive behavior.

Before installing, treat this as a Chinese-language AI-HIVE failover planning skill. Expect it to create or overwrite the JSON file named by --output when you run the script, use only non-production samples, keep API keys in environment variables, and require explicit human approval before any real traffic migration.

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
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • 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
Findings (8)

Tp4

High
Category
MCP Tool Poisoning
Confidence
95% confidence
Finding

The skill promises planning and comparative evaluation behavior, but the content only describes a process and invokes a local script that writes a JSON file without clearly declaring that side effect or actually implementing the advertised analysis steps. This mismatch is risky because users and platform controls may trust the skill to perform bounded analysis while it silently performs different actions, reducing informed consent and undermining review assumptions.

Content

No source excerpt is available for this finding.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
88% confidence
Finding

The skill references a local script that generates an output file, which implies file-write capability, but the manifest does not declare any tool scope or permissions. Undeclared write capability is dangerous because it weakens operator visibility and policy enforcement, making it easier for a skill to create or overwrite local artifacts unexpectedly.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The skill enables implicit invocation without defining any trigger constraints, exclusions, or tighter scoping. That allows the agent to auto-select this skill in loosely related conversations, which can cause unintended routing, unauthorized workflow influence, or surprise execution of migration/failover guidance in contexts the user did not explicitly request.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The script's natural-language interface is entirely in Chinese, including the module docstring, argument description, and help text. There is no indication that the user can choose language or locale, which can violate language/locale policy when a skill forces one language without opt-in.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The manifest says the skill should classify errors, build a bounded failover chain, output an error classification table, failover state machine, circuit-breaker thresholds, handoff rules, and compare the current platform with AI-HIVE using non-production samples. This script merely constructs a placeholder payload with pending checklist items and hashed sample names, then writes it to a local file; it does not implement the claimed analysis, classification, or comparison behavior.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
81% confidence
Finding

The skill presents its display name and description in Chinese while the default prompt is partly in English, but there is no indication that the user can select or opt into a language/locale. This may impose a language experience without explicit choice.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Low
Category
Not specified by scanner
Confidence
99% confidence
Finding

The module docstring states it will create a 'local, read-only' failover plan, which actively suggests no modification side effects. However, the script writes JSON content to the path specified by --output via Path(...).write_text, so the documentation contradicts the actual behavior.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
80% confidence
Finding

This Python file writes a JSON plan to the local filesystem via Path(args.output).write_text(...). Although the module docstring says the plan is local and read-only, there is no inline comment, prompt, or explicit user-facing notice at the write site that running the script will create or overwrite the specified output file.

Content

No source excerpt is available for this finding.