Back to skill

Security audit

BLX Lobster Sticker

Security checks for vulnerabilities and agentic risk

Overview

This is a simple Chinese sticker-selection skill with some over-broad auto-send triggers, but no evidence of hidden code, persistence, credential use, or data exfiltration.

Install this only if you want a Chinese-language assistant to send crayfish sticker images in chat. Consider disabling automatic sends or requiring explicit sticker requests in professional, customer-support, or high-volume conversations because common phrases may trigger stickers unexpectedly.

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
  • 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
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (14)

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The skill description promises automatic sticker sending for a wide range of ordinary emotional expressions, which can cause the agent to trigger on common conversational text without strong intent verification. This is not a code-execution issue, but it can lead to unwanted actions, noisy responses, and reduced user control over agent behavior.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
78% confidence
Finding

The skill defines all trigger phrases, examples, and usage guidance exclusively in Chinese and assumes users will express emotions using those phrases. There is no opt-in for language preference or justification that the skill is intentionally limited to a Chinese-speaking locale, which may violate language/locale policy expectations.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

Many mapped trigger phrases are ambiguous, overlapping, or common in normal conversation, which increases the chance of false activations and incorrect emotional classification. In a messaging assistant, this can create unintended media responses and degrade trust, especially when terms like '牛', '哇', or similar shorthand have multiple meanings.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

Using very common acknowledgements such as '谢谢', '好的', and '行' as direct triggers is especially risky because these appear frequently in ordinary chat and business workflows. This can cause stickers to be sent in inappropriate contexts, including customer support or professional conversations, creating user experience, policy, or reputational issues.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill description says it will automatically send stickers whenever users express a wide range of emotions. In a chat assistant context, this is overly broad behavior that can trigger on normal conversation without explicit user intent, causing unintended actions, spammy responses, or disruption of downstream workflows.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The trigger table includes many ambiguous everyday phrases such as '好的', '为什么', '谢谢', and '不会吧', which commonly appear in ordinary dialogue unrelated to sticker requests. In an auto-triggering skill, these loose keywords increase accidental activation and can cause irrelevant media responses that degrade reliability or interfere with expected assistant behavior.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The instruction to 'automatically send' a matching image whenever the user expresses an emotion is underspecified and lacks permission, rate limiting, or contextual boundaries. That makes the skill prone to unsolicited actions, especially in assistants used for customer support or enterprise chat where automatic media sending may be inappropriate or policy-violating.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
87% confidence
Finding

全文规范、触发词和使用方式均以中文固定表达该技能的交互方式,没有看到允许用户选择语言或说明该技能仅面向特定中文场景的自然语言声明。按照该规则,若技能实际上强制特定语言/locale 而无用户 opt-in 或明确合理限定,属于自然语言政策风险。

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The trigger keywords are very broad and include common conversational terms such as '好', '行', '牛', and emotionally ambiguous phrases, which can cause the skill to activate unintentionally in unrelated contexts. In a chat assistant or customer-service setting, this can lead to incorrect tool use, unwanted image sends, response hijacking from higher-priority tasks, and degraded reliability that an attacker could intentionally provoke with prompt phrasing.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The trigger keywords are very broad and overlap heavily with ordinary conversation terms such as '好', 'OK', '惊讶', '谢谢', and '加油'. In an auto-invocation setting, this can cause unintended skill activation and unsolicited image sending, which may disrupt conversations, override intended assistant behavior, or create spammy/tool-trigger side effects.

Content

No source excerpt is available for this finding.

Vague Triggers

Low
Category
Not specified by scanner
Confidence
88% confidence
Finding

The skill only states that it may refrain from sending when emotion cannot be judged, but it does not define clear negative conditions such as formal contexts, sensitive topics, or repeated trigger suppression. Without explicit boundaries, downstream implementations may over-apply the sticker behavior and create avoidable unintended responses.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Low
Category
Not specified by scanner
Confidence
98% confidence
Finding

L006 states the skill provides '约100+种情绪和物品表情', but the tables only define 50 emotion entries (001-050) and 20 item entries (P01-P20). This is a direct contradiction between the documentation's stated scope and what the file actually specifies.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
84% confidence
Finding

All user-facing trigger phrases and operating instructions are presented only in Chinese, which can force a specific language/locale without offering opt-in or alternatives. The file does not document that the skill is region-specific or limited to Chinese-speaking users.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Low
Category
Not specified by scanner
Confidence
92% confidence
Finding

The manifest says the skill provides 38 emotion stickers and 2 item stickers, but the spec documents about 100+ expressions/items and enumerates 50 emotion entries plus 20 item entries. This is a semantic mismatch about the skill's actual documented scope and behavior.

Content

No source excerpt is available for this finding.