Back to skill

Security audit

应用题数学建模教练

Security checks for vulnerabilities and agentic risk

Overview

ClawScan rates this math word problem coaching skill benign with high confidence: its scans found no executable scripts, while the noted findings concern language requirements, flagged mixed-script content, and examples extending into calculations.

The verdict is benign, and all 12 listed components were scanned. Before installing, consider whether the Chinese-language requirements and examples that go beyond model construction fit your needs; benign does not mean 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
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (7)

The modeling workflow requires Chinese without offering a language choice.

Medium
Category
Output Handling
Confidence
96% confidence
Finding

The instruction “用中文说,不要写公式” explicitly requires the student to describe relationships in Chinese, reinforcing the Chinese-only steps at L087–L088. No language opt-in is provided, and the pedagogical goal of expressing relationships before writing equations does not require a particular natural language.

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.

Worked examples continue into equation solving and numerical answers despite the manifest explicitly limiting the skill to model construction.

Low
Category
Excessive Agency
Confidence
95% confidence
Finding

The manifest explicitly excludes solving equations and performing calculations after a model has been established, delegating that work to xiaozhi-math-problem-solving-coach. However, L39 solves the water-use equation to obtain x = 15, L50 calculates the compound-interest result, and L61 solves the decay model for an age of 17190 years; these teaching examples go beyond identifying quantities and constructing models.

Content

No source excerpt is available for this finding.

Step ② explicitly requires expressing relationships in Chinese without offering a language choice.

Low
Category
Output Handling
Confidence
91% confidence
Finding

The instruction “用中文说关系” specifies Chinese as the language for the modeling step rather than simply asking for a natural-language explanation. Although the reference follows a Chinese mathematics curriculum, it does not explain why this step must use Chinese or allow the learner to choose another language.

Content

No source excerpt is available for this finding.

The extraction workflow requires expressing relationships in Chinese without offering a language choice.

Low
Category
Output Handling
Confidence
90% confidence
Finding

Step ② explicitly says “用中文说关系” (“describe the relationships in Chinese”), imposing a specific output language. Although the resource concerns a Chinese mathematics curriculum, it does not explain why this modeling step requires Chinese or allow the user's preferred language.

Content

No source excerpt is available for this finding.

The reference implements post-model solution steps despite the manifest explicitly limiting the skill to constructing models.

Low
Category
Excessive Agency
Confidence
94% confidence
Finding

The manifest excludes solving and computation after a model is established, and L06 reiterates that boundary. However, L45 goes beyond constructing V(x) and its domain: it differentiates, solves for the critical point, analyzes derivative signs, and computes the maximum volume; the executable assertion at L48 verifies those solution-stage results. Similar completed calculations appear in the repayment, regression, and probability examples.

Content

No source excerpt is available for this finding.

Step 2 explicitly requires Mandarin without offering a language choice.

Low
Category
Output Handling
Confidence
93% confidence
Finding

The instruction “用普通话描述两个量之间的关系” specifies Mandarin for describing the relationship between quantities. It does not offer user opt-in or justify a Mandarin-only constraint; the mathematical modeling task does not inherently require that language.

Content

No source excerpt is available for this finding.