ClawHub
Back to skill

Security audit

统计分析与论文框架 (Agentic AI 科研平台)

Security checks for vulnerabilities and agentic risk

Overview

This skill is a narrow statistics-summary helper with visible local status reporting and a fixed local data-file dependency, but no evidence of hidden exfiltration, persistence, or destructive behavior.

Install only where /home/ubuntu/workspace/demo/mock_data/analysis.json is the intended dataset and localhost:5001/api/report is an expected local reporting service. Verify the provenance of analysis.json before relying on the output for research or medical decisions.

SkillSpector

Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • 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
Findings (4)

Context-Inappropriate Capability

Medium
Confidence
94% confidence
Finding
The skill performs an HTTP POST to a localhost reporting endpoint even though its declared purpose is only statistical analysis and paper-outline generation. Unnecessary network-capable actions expand the skill's privileges and create a path for data leakage, operational signaling, or unintended interaction with local services if the endpoint behavior changes or is attacker-controlled.

Context-Inappropriate Capability

Medium
Confidence
94% confidence
Finding
The completion step repeats the same unnecessary localhost HTTP reporting behavior, again introducing network interaction unrelated to the core analytics function. Repeated non-essential outbound requests increase attack surface and can expose execution metadata or be repurposed for covert signaling.

Vague Triggers

Medium
Confidence
87% confidence
Finding
The trigger description is broad enough to overlap with common requests for statistics or outlining, which can cause the skill to activate outside its intended context. Overbroad activation is dangerous because it can make the agent read fixed local files, generate outputs from unintended data sources, or invoke extra capabilities without clear user intent.

Vague Triggers

Medium
Confidence
85% confidence
Finding
The example phrases are generic and lack constraints or exclusion conditions, making accidental invocation more likely during normal conversation. In this skill, misfire matters because execution includes reading a hard-coded file path and performing reporting actions, so unintended activation can still have security and privacy consequences.

Static analysis

No suspicious patterns detected.