Back to skill

Security audit

品牌视觉一致性审核|AI-HIVE

Security checks for vulnerabilities and agentic risk

Overview

The skill is mostly a disclosed AI-HIVE brand-audit and media-generation workflow, but it handles API credentials and billable generation with weak local safeguards.

Install only if you trust the AI-HIVE workflow and will keep API keys scoped and revocable. Do not use custom or prompt-supplied base URLs, check ~/.ai-hive/config.json before running, start with batch size 1, and manually confirm cost/routing before any generation command.

Vulnerability Patterns
  • 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
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
Findings (2)

T09 · Insecure Skill Coding Practices

Error
Location
scripts/imagegen.py:92
Finding

API bearer credential can be transmitted to an arbitrary configured server

Content
View full analysis
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Warning
Location
scripts/imagegen.py:461
Finding

Unbounded image batch submission can trigger unexpected billable operations

Content
View full analysis
Remediation
View remediation
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Behavioral ASTexec() Call, eval() Call, Dynamic Import
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Findings (21)

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

A skill advertised as returning deviation lists and repair priorities instead appears to expose a general-purpose client for account info, wallet balance, model enumeration, uploads, generation, and task downloads without implementing the promised audit controls. This is dangerous because stakeholders may trust it with sensitive brand assets under the assumption of a constrained reviewer, while the real capability set supports much broader and potentially costly operations.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

A skill advertised as returning deviation lists and repair priorities instead appears to expose a general-purpose client for account info, wallet balance, model enumeration, uploads, generation, and task downloads without implementing the promised audit controls. This is dangerous because stakeholders may trust it with sensitive brand assets under the assumption of a constrained reviewer, while the real capability set supports much broader and potentially costly operations.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

A skill advertised as returning deviation lists and repair priorities instead appears to expose a general-purpose client for account info, wallet balance, model enumeration, uploads, generation, and task downloads without implementing the promised audit controls. This is dangerous because stakeholders may trust it with sensitive brand assets under the assumption of a constrained reviewer, while the real capability set supports much broader and potentially costly operations.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

A skill advertised as returning deviation lists and repair priorities instead appears to expose a general-purpose client for account info, wallet balance, model enumeration, uploads, generation, and task downloads without implementing the promised audit controls. This is dangerous because stakeholders may trust it with sensitive brand assets under the assumption of a constrained reviewer, while the real capability set supports much broader and potentially costly operations.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

The packaged skill is advertised as a brand-visual-consistency audit tool, but the exposed implementation actually provides generic image generation, upload, and task execution flows. This mismatch can mislead operators into granting the skill broader trust and permissions than its stated purpose warrants, creating a capability-confusion risk that can be abused to generate or process arbitrary media outside the intended audit scope.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

The file presents itself as a brand visual consistency audit skill, but the implementation is a generic AI-Hive media/chat client capable of text chat, image generation, video generation, upload, polling, and downloads. This scope mismatch is dangerous because users, reviewers, or policy gates may grant the skill access and trust appropriate for auditing, while it can actually generate arbitrary content and perform unrelated operations.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

The branded entrypoint routes the user into video generation via skill_generate/cmd_video rather than any brand-consistency auditing function. In context, this makes the deception more dangerous because the skill is marketed for compliance-style review, yet operationally serves as a content creation wrapper that could bypass narrower approval expectations.

Content

No source excerpt is available for this finding.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
95% confidence
Finding

The skill documents runnable commands that imply access to environment variables, filesystem, shell, and networked API calls, but it declares no explicit tool scope or permissions boundary. That mismatch can cause an agent runtime to grant broader capabilities than reviewers or users expect, increasing the chance of unintended file access, command execution, secret exposure, or outbound requests.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
86% confidence
Finding

The skill says it uses user brand guidelines as the sole standard for review, but it also invites transforming a reference case into new content with scripts, prompts, code, or task lists. That inconsistency weakens the boundary between auditing and content derivation, which can facilitate IP-sensitive imitation workflows under the cover of a benign brand-check tool.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The documentation for a supposed audit skill includes standalone image and video generation workflows, which broadens the skill from evaluation into content creation. In context, this makes accidental misuse more likely: a reviewer-oriented skill may be invoked in environments where generative actions, external uploads, or billable API calls were not expected or approved.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The natural-language instruction says to produce a "production-ready Chinese workflow," which imposes a specific language on users by default. This is a language/locale policy concern because the file does not offer an alternative language choice or explain why Chinese is required.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
86% confidence
Finding

The natural-language description, argument parser description, skill name, and several metadata values are presented in Chinese with no indication that other languages are supported or optional. This can constitute a language policy violation because the file imposes a specific language context without user opt-in or documented justification.

Content

No source excerpt is available for this finding.

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
70% confidence
Finding

subprocess module calls execute external commands. Without careful input validation, this enables command injection.

Content

Scanner excerpt · scripts/edit_video.py (reported line 20)May include surrounding context.

python
def run(args: list[str]) -> None:
    print(" ".join(args))
    subprocess.run(args, check=True)


def main() -> None:

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The file-level description and CLI branding present the module as a generic AI capability wrapper rather than a narrowly scoped brand-audit skill. In security-sensitive agent ecosystems, inaccurate or overly broad documentation undermines review, increases the chance of unsafe invocation, and hides the true operational surface from users and defenders.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The module docstring and subsequent user-facing prompts/help text are in Chinese, which effectively forces a specific language for users. The file does not offer an opt-in language selection or explain that the skill is intentionally region- or locale-specific.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The packaged skill exposes raw upload and generic task polling commands that are unrelated to a narrow brand-consistency audit workflow. These extra primitives enlarge the attack surface and enable the skill to act as a general-purpose media transport/execution client, which can facilitate unauthorized data movement or unintended downstream model use under the guise of an audit tool.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The module and CLI are documented as a generic AI capability tool, which directly contradicts the branded identity of a brand audit skill. This inconsistent presentation is security-relevant because it obscures true capability during review and can cause operators to misclassify a broad-content-generation tool as a narrow audit utility.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

Querying user-info and wallet balance is unrelated to brand visual consistency auditing and exposes account metadata beyond the stated task. While not directly leading to code execution, it violates least privilege and broadens the skill's access to sensitive operational information that could aid reconnaissance or unintended billing visibility.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

A general-purpose chat_text capability is not justified by a brand-visual-consistency audit skill and materially expands what the skill can do beyond its declared purpose. This unnecessary capability increases the risk of policy circumvention, prompt abuse, and unauthorized use as a generic model access tool under the cover of an audit-branded skill.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
79% confidence
Finding

This code writes a JSON file to a user-specified path, potentially creating directories and overwriting existing content, but the script provides no prior warning, confirmation, or explanatory comment/docstring about this side effect. Aside from printing the resolved path after completion, there is no disclosure before the write occurs.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
93% confidence
Finding

This code file contains natural-language descriptions entirely in Chinese, including the module docstring and later CLI help text, which effectively forces a specific language/locale on users. The policy allows fixed locale only when clearly documented and justified or when users are given a choice, neither of which is present here.

Content

No source excerpt is available for this finding.