Back to skill

Security audit

每周学习复盘

Security checks for vulnerabilities and agentic risk

Overview

ClawScan marks this skill suspicious because a report template directs an extra monthly-summary handoff that conflicts with the stated authorization boundary and could expose student information.

Review the handoff and consent instructions before installing, especially if the skill may handle student information. The scan covered all 10 listed components and completed successfully, but it found a meaningful authorization-boundary issue; the verdict does not establish malicious intent.

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 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
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (1)

The template instructs an additional monthly-summary handoff that the manifest explicitly prohibits.

Medium
Category
Excessive Agency
Confidence
99% confidence
Finding

Section five instructs the agent to hand monthly learning summaries directly to xiaozhi-skill-coordinator, contradicting the manifest’s explicit prohibition on an additional handoff. Following this reference could bypass the intended authorization path, under which the coordinator reads previously authorized growth summaries from learning DNA, and expose a student’s learning weaknesses or behavioral observations without the required consent checks. The legitimate educational context suggests an inconsistent or outdated template rather than malicious intent, but the sensitivity of student records makes the boundary violation meaningful.

Content

No source excerpt is available for this finding.