Back to skill

Security audit

费曼学习法

Security checks for vulnerabilities and agentic risk

Overview

ClawScan rates this skill benign with high confidence: its completed scan found two limited workflow concerns, while the overall assessment found no executable scripts and reported full component coverage.

The scan was complete across all 12 components, and ClawScan’s verdict is benign, not risk-free. Before installing, consider whether you are comfortable with the reported scope and consent concerns: the workflow may direct reteaching and may begin an assessment after a simple acknowledgment.

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
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • 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 assessment workflow directs the skill to teach missing foundations instead of handing instruction off to a subject coach.

Medium
Category
Excessive Agency
Confidence
90% confidence
Finding

The manifest at L003 limits the skill to assessing mastery and locating gaps, with new instruction delegated to the relevant subject coach. However, the outcome table explicitly directs foundational reteaching for grade D at L313, reinforced by the instruction to fill gaps and reteach at L340; this extends the workflow from assessment into instruction without the declared handoff.

Content

No source excerpt is available for this finding.

An ordinary acknowledgment after an AI answer automatically triggers a Feynman assessment.

Medium
Category
Trigger Abuse
Confidence
94% confidence
Finding

The branch starts active verification when a student says “我懂了” (“I understand”) after reading an AI answer. This common acknowledgment does not necessarily express a desire to be tested, and the branch proceeds directly to S1 without requiring agreement.

Content

No source excerpt is available for this finding.