Security audit

Chem Coach AI

Security checks for vulnerabilities and agentic risk

Overview

This is a Markdown-only Chinese chemistry tutoring skill with some broad activation and language-default caveats, but no hidden code, persistence, credential use, or destructive behavior.

Install this as a specialized Chinese Gaokao chemistry tutor. Review the dependent skills separately, especially chem-viz and any configured IMA knowledge base, and consider narrowing trigger keywords if you want it to activate only on explicit tutoring requests.

A.I.G

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

SkillSpector

Vulnerability Patterns

Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
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 (5)

Natural-Language Policy Violations

Medium

Confidence: 82% confidence
Finding: The natural-language description is entirely in Chinese and presents the skill as a Chinese tutoring coach without offering a language choice or opt-in. Under the policy, forcing a specific language or locale without user choice is a reportable natural-language policy issue unless clearly documented as region-specific and justified.

Vague Triggers

Medium

Confidence: 92% confidence
Finding: The invocation guidance lists example phrases but does not define precise trigger boundaries, precedence, or when the skill should not activate. This ambiguity increases the risk of unintended routing from common user utterances, especially in tutoring contexts where phrases like 'I can't do this problem' are routine and could silently force a specific mode.

Vague Triggers

Medium

Confidence: 93% confidence
Finding: The example trigger phrase is very close to ordinary student chat and provides no strong boundary that this is an explicit skill invocation rather than normal conversation. In an agent environment, this can cause accidental activation, changing response behavior unexpectedly and potentially bypassing higher-priority safety or routing expectations.

Natural-Language Policy Violations

Medium

Confidence: 87% confidence
Finding: The skill description and behavior are written to operate in Chinese without offering language negotiation or explicit opt-in, which can cause it to take over the interaction in a language the user did not request. This is primarily a safety and usability issue rather than a direct exploit, but it can mis-handle user intent, reduce transparency, and increase the likelihood of confusing or inappropriate responses when auto-invoked.

Vague Triggers

Medium

Confidence: 92% confidence
Finding: The trigger keyword list includes broad, high-frequency terms such as “化学”, “实验”, and “有机”, which can cause the skill to activate in contexts far beyond the intended gaokao chemistry tutoring use case. Over-broad activation can override user intent, surface specialized instructional behavior unexpectedly, and increase the chance of unwanted tool invocation or disclosure of domain-specific guidance in unrelated conversations.

Static analysis

No suspicious patterns detected.