Back to skill

Security audit

鸡尾酒顾问 cocktail-advisor

Security checks for vulnerabilities and agentic risk

Overview

This is a cocktail recommendation skill with some disclosed low-risk search and preference-memory behavior, but no evidence of malicious or hidden high-impact actions.

Before installing, expect a Chinese-focused cocktail advisor that may search the web for recipe gaps and may save cocktail preferences in a workspace memory file. Use it if that personalization is acceptable, and avoid putting sensitive personal details in feedback unless you are comfortable retaining them locally.

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
Vulnerability Patterns
  • System Prompt LeakageDirect Leakage, Indirect Extraction, Tool-Based Exfiltration
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Findings (6)

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

The manifest description and all operational instructions are in Chinese, and the output template is defined only in Chinese, with no indication that the user may choose another language. This appears to impose a language preference without opt-in or a documented regional justification.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The activation conditions are broad enough to trigger on ordinary phrases about drinks, ingredients, or scenarios, which can cause unintended invocation. Unintended activation matters more here because the skill also includes external search and persistent memory behaviors, increasing the chance of surprise data handling.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The documented behavior includes real-time external searches even though the manifest presents the skill as an internal cocktail advisor. Undisclosed outbound retrieval can expose user-provided ingredients or preferences to third parties and changes the trust boundary of the skill.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill’s manifest and primary purpose are cocktail recommendations, but it also instructs the agent to persist user preference data in a workspace file. This expands behavior from transient advice into stateful data handling, creating privacy and scope-creep risk that is not disclosed in the manifest or clearly consented to by the user.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The skill directs persistent tracking of user preferences in a workspace file without warning the user or obtaining consent. Even if the data seems low sensitivity, preference logs can accumulate behavioral information over time and create avoidable privacy and retention risks.

Content

No source excerpt is available for this finding.

Ssd 3

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

Natural-language feedback storage in workspace memory creates data retention risk because free-text comments may include more personal information than intended. Over time, this can lead to accumulation of unnecessary personal data, prompt leakage across sessions, or reuse outside the user’s expectations.

Content

No source excerpt is available for this finding.