Back to skill

Security audit

物理建模教练

Security checks for vulnerabilities and agentic risk

Overview

ClawScan rates this skill benign with high confidence: scans found no executable files or AIG issues, while Skillspector flagged limited scope inconsistencies in its physics coaching instructions.

The verdict is benign, and all 15 reported components were fully inspected. Before installing, consider whether the skill’s coaching scope inconsistencies matter for your use; the verdict does not mean the skill is risk-free.

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 (10)

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.

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.

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 modeling workflow includes numerical substitution and solving, which the manifest explicitly delegates to the separate problem-solving coach.

Medium
Category
Excessive Agency
Confidence
98% confidence
Finding

The workflow directs numerical substitution and solving even though the skill description delegates complete solution procedures and calculation to another coach. This creates a bounded routing and scope-integrity weakness: the modeling coach may proceed into calculations instead of stopping at model selection and equation construction, although limited illustrative calculation is not inherently unsafe.

Content

No source excerpt is available for this finding.

The reference instructs the coach to decline high-school modeling instruction despite the manifest explicitly including high-school compulsory and elective-compulsory physics.

Medium
Category
Excessive Agency
Confidence
99% confidence
Finding

The reference explicitly directs the coach not to explain or assign high-school material, contradicting the declared coverage of compulsory and elective-compulsory high-school physics. Following this reference could systematically deny supported instruction and omit relevant learning records; this is a functional instruction-integrity issue rather than evidence of malicious compromise.

Content

No source excerpt is available for this finding.

The shared escalation rules direct this modeling-only skill to provide complete solutions instead of handing off to the physics problem-solving coach.

Medium
Category
Excessive Agency
Confidence
93% confidence
Finding

The manifest explicitly excludes complete solution procedures and calculations, assigning those to the physics problem-solving coach. However, L27 defines L6 as explaining the original problem's complete solution, and L35 directs the skill to enter L6 after repeated requests for direct explanation, without preserving the manifest's handoff requirement.

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 modeling workflow includes calculation and solving unknowns despite the manifest explicitly excluding those tasks.

Low
Category
Excessive Agency
Confidence
97% confidence
Finding

The manifest at L003 excludes complete problem-solving workflows and calculation, delegating them to the physics problem-solving coach. However, L123–L126 instruct numerical substitution and solving unknowns, and L358 further directs simultaneous equation solving, extending the workflow beyond model selection and mathematical formulation.

Content

No source excerpt is available for this finding.