Back to skill

Security audit

Ai Reviewer

Security checks for vulnerabilities and agentic risk

Overview

The skill is not malicious, but it presents itself as an image quality gate while its code does not actually inspect images and deterministically approves them.

Install only if you treat this as a prototype or workflow stub. It should not be trusted to approve ecommerce images until real image analysis is implemented, and you should review the local history logging behavior if review metadata is sensitive.

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
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Findings (11)

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
80% confidence
Finding

The skill advertises no explicit tool scope, yet the surrounding project description indicates code capabilities that may read and write files. When a skill lacks declared permissions, operators and orchestrators cannot reliably constrain what it is allowed to access, which can lead to unintended file access or modification if the implementation is invoked with broader defaults.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The input configuration explicitly sets language: "zh-CN", requiring a specific locale without offering user choice or opt-in. This is a natural-language policy issue because the file mandates a single language rather than documenting it as optional or context-specific.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The config section hard-codes language: "zh-CN", which indicates the skill is designed to operate in a single locale without presenting this as a user choice. The file does not explain a region-specific compliance reason or provide an opt-in mechanism.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

Describing the skill as 'always on' creates unclear activation boundaries and suggests it may monitor or intervene without an explicit user request. Ambiguous activation state increases the chance of overreach, accidental processing of unrelated inputs, and unexpected access to project artifacts in multi-skill environments.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The natural-language trigger example is so broad that ordinary conversation could unintentionally activate the skill. In an agent system, ambiguous activation can cause the reviewer workflow to run on user content or files without a clear, deliberate request, increasing the risk of unauthorized data handling or confusing tool execution.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The default configuration sets the language to "zh-CN", and the surrounding docstrings/messages indicate the skill operates in Chinese by default. This enforces a specific language/locale choice without offering the user a documented opt-in or alternative, which matches the language/locale policy violation criteria.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The agent presents itself as performing image quality review, but the implementation does not inspect image content and instead returns deterministic placeholder scores. In a workflow that relies on this review to gate output quality, this can cause unsafe or noncompliant images to be approved, creating integrity and trust failures.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The configuration hard-codes "language": "zh-CN", which imposes a specific language/locale setting. Under the policy, locale constraints should either be user-selectable or clearly justified as region-specific; this file provides neither.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The manifest specifies language: zh-CN in the input contract and again in config, which indicates the skill is constrained to Chinese. There is no indication of user opt-in, alternative locale support, or justification that this is a region-specific skill, so this is a natural-language locale policy concern.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Low
Category
Not specified by scanner
Confidence
92% confidence
Finding

The comments explicitly admit the core checking logic is a simplified placeholder, yet the surrounding API and agent description frame it as a real quality-control component. This mismatch increases the risk that operators deploy it as a trusted reviewer, leading to false approvals and weakened assurance.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
89% confidence
Finding

The description string '质检规则库 - 质检师专用' is natural-language content indicating the skill is intended specifically for Chinese-speaking quality inspectors. Because this file provides no opt-in, alternative language support, or documented region-specific justification, it may violate the language/locale policy for user choice.

Content

No source excerpt is available for this finding.