Back to skill

Security audit

数学错误DNA

Security checks for vulnerabilities and agentic risk

Overview

ClawScan rates this skill benign with medium confidence: the scanners report a scope mismatch around progressive math practice, while finding no AIG issues; Skillspector coverage was partial.

The verdict is benign, but confidence is medium and the scan was partial because two local references could not be resolved. Before installing, consider whether the progressive-practice scope mismatch is acceptable and whether you are comfortable with that incomplete evidence.

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
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (4)

The skill directly delivers progressive exercises despite explicitly delegating tiered practice to another skill.

Medium
Category
Excessive Agency
Confidence
97% confidence
Finding

The manifest at L003 excludes tiered practice and directs it to xiaozhi-math-gradient-trainer. However, L353–L356 instruct this skill to generate and administer three progressively structured exercises, and L179–L180 similarly prescribe progressive interference-condition exercises. Unlike a minimal diagnostic question used to identify an error subtype, these sequences implement the training capability the manifest expressly excludes.

Content

No source excerpt is available for this finding.

Ae4

Medium
Category
analysis-evasion
Confidence
80% confidence
Finding

Suspicious Unicode normalization or mixed-script content

Content

No source excerpt is available for this finding.

Ae4

Medium
Category
analysis-evasion
Confidence
80% confidence
Finding

Suspicious Unicode normalization or mixed-script content

Content

No source excerpt is available for this finding.

The reference workflow assigns progressive practice to this skill even though the manifest explicitly delegates that activity to xiaozhi-math-gradient-trainer.

Medium
Category
Excessive Agency
Confidence
97% confidence
Finding

The reference workflow assigns progressive exercise generation and execution to this skill, whereas the metadata explicitly delegates layered practice to xiaozhi-math-gradient-trainer. This conflicting operational guidance can cause an agent to exceed its declared scope and bypass the intended specialist handoff, although the supplied content does not demonstrate data disclosure, privilege escalation, or destructive actions.

Content

No source excerpt is available for this finding.