Security audit

AI科研全流程助理

Security checks for vulnerabilities and agentic risk

Overview

This research-assistant skill is coherent and purpose-aligned, with disclosed report generation and web-search-based academic workflows, but users should be aware it can write an HTML report file and may activate on broad academic phrases.

Install if you want a broad academic assistant that can search the web, process research text or uploaded data, and generate a local HTML report. Review the generated report before relying on it, manually verify citations and factual claims, avoid sensitive unredacted research data, and keep report output paths in a normal project or downloads folder.

SkillSpector

Vulnerability Patterns

Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access

Findings (4)

Lp3

Medium

Category: MCP Least Privilege
Confidence: 90% confidence
Finding: The skill instructs the agent to invoke a local Python script that writes an HTML report to the current working directory, but the manifest declares no permissions or safeguards around file-writing behavior. This mismatch can lead to unreviewed filesystem side effects, confusion in sandboxing/policy enforcement, and potentially unsafe writes if output paths or JSON content are influenced by user input.

Vague Triggers

Medium

Confidence: 89% confidence
Finding: The README shows very broad trigger phrases such as '科研助手', '学术助理', '论文写作', and 'literature review', which can cause accidental or overly permissive activation in unrelated contexts. In an agent skill that can perform web search, generate reports, and influence research workflows, loose activation boundaries increase the chance of unintended invocation, prompt confusion, and misuse of the skill outside the user's intended scope.

Vague Triggers

Medium

Confidence: 84% confidence
Finding: Several trigger phrases are broad everyday academic requests such as writing papers, polishing text, experiment design, and data analysis, which can cause the skill to activate in contexts where the user did not explicitly intend this specific skill. Over-broad activation increases the chance of accidental tool use, unintended processing of uploaded content, or surprising behaviors such as report generation and data handling.

Vague Triggers

Medium

Confidence: 80% confidence
Finding: The quick-start section says the skill will 'automatically recognize' four input modes and 'automatically route' to modules, but it does not define clear boundaries for when activation should occur or when user confirmation is required. That ambiguity can cause unintended triggering on ordinary conversation, uploads, or partial academic requests, leading to unnecessary data processing and unexpected actions.

Static analysis

No suspicious patterns detected.