Back to skill

Security audit

康奈尔笔记

Security checks for vulnerabilities and agentic risk

Overview

ClawScan rates this skill suspicious because it reports student response tracking and permits broader student note writeback than its stated retrieval-only scope suggests.

Consider whether student response tracking and broader note writeback are appropriate before installing. The scan reports complete inspection of the listed components, and these concerns support the suspicious Review verdict; they do 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 (2)

The skill persists student-response metrics and produces effectiveness analysis beyond the manifest’s retrieval-only usage report.

Medium
Category
Excessive Agency
Confidence
94% confidence
Finding

The manifest limits usage reporting to which notes were retrieved, but Module E additionally persists response counters and reports whether prompted knowledge was used. This expands behavioral tracking of students beyond the advertised scope; profile opt-in and the exclusion of conversation content reduce exposure but do not clearly authorize this additional tracking.

Content

No source excerpt is available for this finding.

Learning-DNA writeback permits arbitrary note data instead of only the two declared summaries.

Medium
Category
Excessive Agency
Confidence
97% confidence
Finding

The extensionPatch wrapper restricts writes to notes, but notes itself accepts arbitrary properties and nested data, including full note text rather than only the declared summaries and update timestamp. If schema validation is relied upon as the writeback boundary, an erroneous or manipulated payload can pass validation and share student information beyond the authorized summary scope; general cross-skill consent does not enforce data minimization.

Content

No source excerpt is available for this finding.