Back to skill

Security audit

AI发型发色预览图|AI-HIVE

Security checks for vulnerabilities and agentic risk

Overview

This appears to be a real AI-HIVE hairstyle preview helper, but it needs review because its API client can send API keys and user-selected media to configurable or unvalidated endpoints.

Install only if you trust AI-HIVE and are comfortable uploading authorized portrait/media files to its service. Use the default official HTTPS endpoint, avoid custom --base-url or AI_HIVE_BASE_URL values, protect ~/.ai-hive/config.json, and confirm prompts, routing, and cost before generation.

Vulnerability Patterns
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • 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
Findings (2)

T09 · Insecure Skill Coding Practices

Warning
Location
scripts/imagegen.py:91
Finding

Configurable API Endpoint Can Expose Credentials and Private Media

Content
View full analysis
Remediation
View remediation

T08 · Insecure Dependencies

Note
Location
scripts/imagegen.py:7
Finding

Unpinned Third-Party Dependency Installation

Content
View full analysis
Remediation
View remediation
Vulnerability Patterns
  • 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 Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Findings (14)

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

A specialized media-editing skill that also exposes generic chat, video generation, account information access, model enumeration, and local API-key setup has a materially expanded capability surface beyond its declared role. That discrepancy can mislead security review and cause the skill to be invoked in contexts where it can access sensitive account metadata or perform unrelated actions, raising the risk of abuse and unintended data exposure.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
93% confidence
Finding

A specialized media-editing skill that also exposes generic chat, video generation, account information access, model enumeration, and local API-key setup has a materially expanded capability surface beyond its declared role. That discrepancy can mislead security review and cause the skill to be invoked in contexts where it can access sensitive account metadata or perform unrelated actions, raising the risk of abuse and unintended data exposure.

Content

No source excerpt is available for this finding.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
89% confidence
Finding

The skill advertises executable workflows that rely on environment variables, local files, shell commands, and outbound network access, but it does not declare any explicit tool scope or permission boundaries. That creates unnecessary ambient authority: if the runtime grants broad tools by default, the skill could read secrets, write files, or make external requests without a narrowly documented contract, increasing the blast radius of prompt injection or operator error.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

Activation on a common, underspecified phrase like a generic hairstyle-related term can cause accidental invocation. While not inherently malicious, this raises the probability of misrouting user requests into a skill that may involve external services or more intrusive processing than the user intended.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The trigger condition includes broad marketing and content-production requests, not just hairstyle/hair-color preview intents. Overbroad activation increases the chance the skill is selected for unrelated user requests, which can expose users to unnecessary external API usage, file/network operations, or outputs outside the expected domain.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The default prompt hard-codes Chinese output regardless of user preference, which can cause consent and clarity issues if the user expects another language. While not directly enabling code execution or data exfiltration, it can degrade user understanding of generated workflows, API commands, and safety limitations, increasing the risk of misuse or mistaken approval.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The manifest enables implicit invocation, which allows the skill to be triggered without an explicit user request for this specific capability. In a skill that can generate runnable workflows and AI-HIVE commands for image-editing tasks, broad implicit activation increases the chance of unintended use, including processing sensitive portrait-editing requests or policy-sensitive transformations without sufficient user confirmation.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The natural-language strings in the module docstring and argument parser description specify the skill name entirely in Chinese, and the generated metadata continues that language choice. Because the file provides no user opt-in, language selection, or justification that this is a China-specific skill, it appears to impose a specific language/locale by default.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The file is presented as a hairstyle/color preview skill, but the module-level implementation is a broad AI Hive client supporting chat, image, video, model enumeration, uploads, and account queries. This scope mismatch increases the attack surface and can enable capability abuse if the skill is invoked in environments that assume it is limited to hairstyle-preview image editing.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

This Python file contains natural-language docstrings, prompts, help text, and status messages entirely in Chinese, starting with the module description and continuing throughout the CLI. Under the policy, forcing a specific language without user opt-in is a natural-language policy violation unless a justified locale restriction is documented, which is not present here.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

Text chat capability is unrelated to the stated hairstyle-preview purpose and allows arbitrary prompt submission to a general text model. In a skill ecosystem, this can be repurposed as an unsanctioned generic LLM access path, bypassing policy assumptions that this skill only performs constrained image-preview tasks.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

Full video-generation support materially exceeds the hairstyle/color preview image workflow and enables transformation of uploaded media into arbitrary videos. Because this skill handles user-supplied media and remote generation/download, the extra modality broadens misuse potential, cost exposure, and policy evasion risk under a misleading beauty-preview label.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
82% confidence
Finding

The entire natural-language description and usage guidance are written as a Chinese-only skill experience, and there is no statement offering language selection or opt-in for another locale. Under the policy, forcing a specific language without user choice can be a locale-policy issue unless explicitly documented and justified.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Low
Category
Not specified by scanner
Confidence
85% confidence
Finding

User-info and model-enumeration operations expose account metadata, wallet information, and platform capabilities that are not required for end-user hairstyle preview generation. While not an exploit by themselves, they create unnecessary information disclosure and reconnaissance surface if the skill is accessible to broader callers.

Content

No source excerpt is available for this finding.