Back to skill

Security audit

作业跟进管家

Security checks for vulnerabilities and agentic risk

Overview

ClawScan rates this skill benign with medium confidence: the scanners reported two moderate scope mismatches, while the completed AIG scan found no issues; Skillspector coverage was partial.

Consider whether you are comfortable with those two reported scope mismatches before installing. The verdict is benign, not risk-free; Skillspector marked its analysis partial because three local references could not be resolved, so the 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
  • 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 (2)

The workflow permits generating verification questions despite the manifest explicitly excluding question generation.

Medium
Category
Excessive Agency
Confidence
98% confidence
Finding

The manifest at L003 says this skill does not generate questions and routes classroom exercise creation to exam-designer. However, L263 instructs this skill to self-check and output AI-generated analogous or simplified verification questions. Requiring teacher validation before storage or student use does not resolve the explicit scope mismatch.

Content

No source excerpt is available for this finding.

The schema directs ongoing overdue tracking for unfinished and unsubmitted work despite the manifest’s explicit prohibition on tracking non-submission.

Medium
Category
Excessive Agency
Confidence
91% confidence
Finding

The manifest explicitly states “不追未交、不催交” (do not track unsubmitted work or chase submissions). In contrast, L278 instructs readers to calculate overdue days through the current date for work marked 已布置、部分提交、未交, establishing ongoing lateness tracking rather than merely recording completed classroom practice or reviewing schoolwork.

Content

No source excerpt is available for this finding.