Back to skill

Security audit

把书和大佬的方法蒸成AI技能

Security checks for vulnerabilities and agentic risk

Overview

The scanners classify this skill as benign, with notes about broader task delegation and required Chinese trigger phrases; the review is incomplete because three referenced paths were not found in the scan.

The verdict is benign, but it is not a guarantee of safety. Before installing, consider whether you want a skill that may coordinate agents to complete broader tasks and require Chinese trigger phrases; the scan also could not resolve three referenced paths, so its evidence is incomplete.

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
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • 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)

The skill expands from producing reusable skills into orchestrating agents to execute the user's underlying task.

Medium
Category
Excessive Agency
Confidence
90% confidence
Finding

The manifest describes distilling sources into installable skills, checklists, and procedures, with goal-based recruitment used to select suitable sources. L104 and L108–L114 instead direct the assistant to decompose and complete substantial user tasks through delegated agents, including dispatching work and accepting deliverables. Executing the underlying task is a distinct operational capability, not merely generating or validating a distilled skill.

Content

No source excerpt is available for this finding.

Generated skills must include Chinese trigger phrases without user language opt-in.

Medium
Category
Output Handling
Confidence
97% confidence
Finding

L198 requires Chinese trigger phrases in the generated skill description, and L208 makes their inclusion a mandatory acceptance criterion. This imposes a specific output language without offering a user choice or documenting a locale-specific justification.

Content

No source excerpt is available for this finding.