Back to skill

Security audit

重大疾病距离模拟器

Security checks for vulnerabilities and agentic risk

Overview

This skill is a disclosed health-risk simulator, but it collects and uses sensitive health profiles with external searches and unvalidated medical-style predictions without enough consent, privacy, or safety boundaries.

Install only if users will be clearly told before use that sensitive health details may be collected and used in online searches. The skill should require explicit consent, avoid sending detailed personal combinations to search providers, keep profile data session-limited unless the user opts in, and frame results as general education rather than individual prognosis or medical urgency.

Vulnerability Patterns
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • 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
Findings (2)

other

Warning
Location
SKILL.md:42
Finding

Sensitive Health Profile Exposure Through External Search Queries

Content
View full analysis
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Error
Location
SKILL.md:157
Finding

Unvalidated Personalized Disease-Onset Model and Unsupported Medical Recommendations

Content
View full analysis
Remediation
View remediation
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (5)

Missing User Warnings

High
Category
Not specified by scanner
Confidence
96% confidence
Finding

The skill is designed to collect sensitive health data, family medical history, body metrics, lifestyle habits, and location before presenting any clear privacy notice, consent flow, or explanation of external transmission. This is dangerous because users may reveal regulated or highly personal information without understanding retention, sharing, or that queries may be sent to networked RAG/search systems.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The trigger phrases are broad enough that ordinary discussion about health risk or disease could activate the skill without clear user intent. In this skill's context, unintended activation is more dangerous because activation leads into collection of highly sensitive health, family-history, and lifestyle data and may also trigger external RAG lookups.

Content

No source excerpt is available for this finding.

Ssd 3

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The instruction to 'directly record' sensitive health information during conversation encourages accumulation of a detailed health profile without any stated data minimization, retention limits, or consent boundary. In a health-risk skill, persistent recording raises the chance of privacy harm, unauthorized reuse, and over-collection beyond what is needed for a single response.

Content

No source excerpt is available for this finding.

Ssd 3

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

State tracking across multiple turns, and potentially across sessions, creates a cumulative sensitive dossier containing health, family-history, behavioral, and geographic information. This becomes more dangerous in this skill because the entire product value depends on longitudinal profile enrichment, which increases exposure if logs, memory stores, or downstream tools are accessed improperly.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Low
Category
Not specified by scanner
Confidence
96% confidence
Finding

在L006与L059-L061、L343中,技能反复声明输出是“2段式教育报告”,仅包含距离可视化和身体信号+缓解行动清单。但L188明确写成“输出:3段式教育报告”,与前述意图说明直接冲突,造成技能实际宣称的交付形态不一致。

Content

No source excerpt is available for this finding.