Back to skill

Security audit

小沐宝咖啡

Security checks for vulnerabilities and agentic risk

Overview

The skill is a coherent coffee-stall helper, but it under-scopes child safety and privacy around location, social posts, photos, and contact details.

Review before installing. The skill appears intended for a family coffee-stall workflow and does not contain executable malware, but it should be tightened before public use: avoid precise public location schedules for a child, require adult review for any post with time/location/photos, use an adult-managed public channel for updates, do not collect customer contact details informally, and narrow triggers to brand-specific terms.

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
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Findings (17)

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The skill explicitly instructs the agent to answer in a fixed '7-year-old girl' persona, which can override normal assistant behavior without clear user opt-in. This is primarily a safety and UX issue: users may receive role-played responses when they expected neutral assistance, and the child persona can increase persuasion or emotional trust in ways users did not request.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The auxiliary trigger list contains broad everyday terms such as '菜单', '摆摊', '咖啡摊', and '拉花', which can match many unrelated conversations and cause unintended activation. In practice this can hijack normal chat flow, inject the skill's branded instructions unexpectedly, and increase the chance that persona constraints or other skill behaviors override user intent.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The auxiliary trigger list includes broad everyday coffee-related terms such as '拉花' and '手磨咖啡', which can cause the skill to activate in conversations that are not actually about this specific brand. This creates prompt-routing risk: unrelated user queries may be handled by the wrong skill, leading to confusing responses, brand/persona leakage, or reduced reliability of downstream actions.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

The skill enforces a fixed childlike Chinese persona and specific forms of address without accommodating the user's language, tone, or context. While not a direct security exploit, this can cause inappropriate or manipulative responses in unrelated settings, and increases the risk of unsafe impersonation or trust-shaping if the skill is triggered accidentally.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The template explicitly prompts publishing a child's current time and location for a mobile coffee stall, which creates a real-time stalking and child-safety risk. In this skill's context, the danger is elevated because the business is operated by a 7-year-old girl, making precise schedule/location disclosure especially sensitive.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

This content suggests collecting customer contact information without any privacy notice, consent language, or instruction that only a guardian should handle such data. Because the operator is explicitly a 7-year-old child, the absence of adult-controlled collection and safeguarding makes the privacy issue more serious than in an ordinary retail script.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The script invites customers to send photos or videos to the child's father without mentioning consent boundaries, appropriate sharing methods, or risks of distributing media involving a minor. This can normalize informal sharing of child images and create privacy and safety issues for the child and the family.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The document publishes a minor's specific recurring times and likely locations for future appearances, including neighborhood-level details and a history of past outings. Because the skill is about a 7-year-old girl's mobile coffee stall, this materially increases stalking, harassment, and physical safety risk; the child-focused context makes the exposure significantly more dangerous than for an adult-run business.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The schedule tool is explicitly designed to return future date, location, and operating time data for a mobile coffee stall run by a 7-year-old child. Publishing or programmatically exposing precise whereabouts and timing for a minor materially increases stalking, harassment, and physical safety risks, and the family-oriented context makes this more sensitive than an ordinary business schedule feed.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
89% confidence
Finding

The manifest defines the assistant persona, tone rules, and example responses entirely in Chinese, including mandatory phrasing such as using specific Chinese kinship terms for customers. This indicates a fixed language/locale behavior without offering the user a choice or documenting a justified region-specific constraint.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
84% confidence
Finding

This markdown file contains user-facing instructions exclusively in Chinese, and there is no note that the skill is intended only for Chinese-speaking users or that another language option is available. Under the natural-language policy rule, forcing a specific language without user opt-in can be a locale policy concern.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
91% confidence
Finding

This markdown file presents all user-facing content exclusively in Chinese, which can violate a language/locale policy when no user opt-in or alternative language is offered. The file does not indicate that the skill is region-specific or that Chinese is an intentional, documented locale constraint.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
86% confidence
Finding

The template encourages sharing stories about identifiable customer interactions and suggests including people-related narrative content, which can expose third parties without consent. Although it recommends blurred back-view images, there is no explicit requirement to anonymize details, obtain consent, or avoid recognizable descriptions, creating avoidable privacy risk.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Low
Category
Not specified by scanner
Confidence
88% confidence
Finding

The template encourages asking customers to leave contact information for future stall notifications, but it provides no data-minimization, consent, retention, or guardian-handling guidance. In the context of a child-run coffee stall, collecting personal contact details creates unnecessary privacy risk and can expose both customers and the child/family to misuse or oversharing.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
75% confidence
Finding

Referring customers to the father's social feed for updates can expose personal social account information and create ongoing contact expectations. The markdown lacks any warning that a dedicated public or business-facing account should be used instead of a personal account.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
94% confidence
Finding

The document uses Chinese throughout for headings, instructions, and status labels, which imposes a specific language on readers without any opt-in or stated regional justification. Under the policy rule for language or locale constraints, this is a natural-language policy concern because no alternative language option is provided.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Low
Category
Not specified by scanner
Confidence
72% confidence
Finding

generate_moments_copy creates social-media posts for Moments/朋友圈, extending the skill from answering coffee-stall questions into marketing content generation. While loosely related to operations, that capability is not an obvious requirement of a coffee information lookup skill and is especially sensitive given the branding centers on a 7-year-old child.

Content

No source excerpt is available for this finding.