Security audit

benzhi

Security checks for vulnerabilities and agentic risk

Overview

The skill is a coherent analysis prompt, but it asks the agent to retain and reuse user analysis history without clear limits or consent.

Review this skill before installing if your analyses may include private, sensitive, or client-specific topics. It should be adjusted so analysis history is not stored or reused unless the user explicitly opts in and retention limits are clear.

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

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
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)

Natural-Language Policy Violations

Medium

Confidence: 84% confidence
Finding: The manifest description and the full instructions are written entirely in Chinese, and the skill presents Chinese trigger phrases and output labels without indicating that users may choose another language. This can amount to an implicit language policy restriction when no opt-in or justified locale scope is documented.

Vague Triggers

Medium

Confidence: 96% confidence
Finding: The trigger phrase "为什么XX？" is extremely broad and can match ordinary conversation, causing the skill to activate outside its intended scope. Over-broad activation can hijack unrelated user requests, leading to prompt misrouting, degraded behavior, and unexpected handling of sensitive content under this skill’s framework.

Ssd 3

Medium

Confidence: 95% confidence
Finding: The self-learning and passive logging sections instruct retention of analyzed user content, conclusions, counter-examples, and meta-insights across analyses, which creates a cross-session data retention and privacy risk. If implemented, this can store sensitive user inputs without consent, enable profiling, and leak prior users’ information into later sessions or model updates.

Vague Triggers

Low

Confidence: 89% confidence
Finding: The automatic quick/deep mode selection rule is underspecified, so different agents may make inconsistent choices for the same input. Ambiguous routing logic increases unpredictability and can lead to excessive data processing, unnecessary long-form reasoning, or failure to respect user intent.

Static analysis

No suspicious patterns detected.