Back to skill

Security audit

拍照识万物

Security checks for vulnerabilities and agentic risk

Overview

This is a coherent image-identification skill with some usability and privacy-notice gaps, but no hidden execution, persistence, credential access, or destructive behavior was found.

Install this if you want a Chinese-language image-recognition assistant. Avoid uploading sensitive photos such as IDs, private documents, faces, or confidential workplace images unless you are comfortable having them processed by the model. Be aware that broad phrases may invoke the skill when an image is present, so review the result if routing seems unexpected.

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
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (8)

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
86% confidence
Finding

All user-facing instructions and trigger phrases are presented only in Chinese, with no indication that other languages are supported or that the user can opt into Chinese. This can constitute a language-policy issue when a skill implicitly constrains interaction to a specific language without explicit choice or justification.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The trigger phrases are broad, generic requests such as '这是什么' and '帮我看看', which can overlap with many normal conversations and cause the skill to activate unintentionally. In a vision-analysis skill, this can lead to inappropriate routing of user requests, unexpected image processing, and confusing or privacy-impacting behavior if photos are analyzed when the user did not explicitly intend to invoke this skill.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

The natural-language description and prescribed response patterns are written to require Chinese output conventions such as "用老百姓日常叫法" without offering any language choice. This can violate language or locale policy where user-facing skills should not force a specific language absent opt-in or documented regional justification.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The trigger set is broad enough to activate on common phrases such as '这是什么' or '帮我看看', which are widely used outside image-identification contexts. This can cause unintended skill invocation, hijacking unrelated user requests and increasing the chance of wrong tool routing or unexpected handling of user data.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
98% confidence
Finding

The manifest enumerates multiple ambiguous triggers, including short or generic phrases that can match normal conversation without clear intent to use this skill. In an agent ecosystem, this raises the risk of accidental activation, interference with other skills, and unintended processing of user inputs.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The document states that user-uploaded images are automatically sent to the model, but it provides no warning about external transmission, retention, or privacy risks. Because images may contain faces, documents, location data, or other sensitive information, this omission can cause users to disclose personal or confidential data without informed consent.

Content

No source excerpt is available for this finding.

Overly Broad Trigger

Low
Category
Trigger Abuse
Confidence
89% confidence
Finding

The trigger '识图' is extremely short and semantically broad, making false matches more likely during normal conversation or when users mention image-related topics casually. Short ambiguous triggers degrade intent precision and can lead to accidental invocation of this skill over more appropriate handlers.

Content

No source excerpt is available for this finding.

Overly Broad Trigger

Low
Category
Trigger Abuse
Confidence
88% confidence
Finding

The trigger '识物' is too short and underspecified to safely represent a unique intent. Its ambiguity increases the probability of unintended routing, especially in systems that match partial or fuzzy user utterances.

Content

No source excerpt is available for this finding.