Back to skill

Security audit

防止AI幻觉

Security checks for vulnerabilities and agentic risk

Overview

This skill is a markdown-only anti-hallucination checklist with one disclosed but loosely scoped memory note.

Before installing, consider whether you want a skill that may ask the agent to write mistake records into memory. A safer use pattern is to store only brief, non-sensitive corrections and avoid recording private task details.

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
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • 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 (3)

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
98% confidence
Finding

The entire skill description and operating instructions are written only in Chinese, and there is no statement that use of Chinese is optional or limited to a China-specific context. Under the language/locale policy rule, this can constitute a locale constraint without user opt-in.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Low
Category
Not specified by scanner
Confidence
84% confidence
Finding

The manifest describes a graded checking mechanism focused on preventing unsupported factual output and marking uncertainty. However, the documented workflow adds a separate behavior of persisting mistakes to memory, which goes beyond output-checking and enters stateful incident retention.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Low
Category
Not specified by scanner
Confidence
80% confidence
Finding

A skill whose stated purpose is to enforce evidence-based answering and confidence labeling would be expected to define checking and response rules. Persisting information to memory is a distinct capability that is not obviously required by the manifest description and could retain broader contextual data than necessary.

Content

No source excerpt is available for this finding.