Back to skill

Security audit

AI电商上架套图|AI-HIVE

Security checks for vulnerabilities and agentic risk

Overview

The skill mostly matches an AI-HIVE ecommerce image workflow, but it handles API keys and uploads through a user-configurable endpoint that could expose credentials or selected media if misconfigured.

Review before installing. Use only the default AI-HIVE endpoint unless you fully trust a custom endpoint, keep the API key scoped and revocable, avoid setting AI_HIVE_BASE_URL from untrusted environments, and upload only media you are comfortable sending to AI-HIVE. Non-Chinese users should expect Chinese-first prompts and output.

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

Unrestricted API Base URL Can Expose Bearer Credentials and Uploaded 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
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • 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 (12)

Tp4

High
Category
MCP Tool Poisoning
Confidence
95% confidence
Finding

This finding indicates both over-claiming and undeclared capability expansion, including general chat, video generation, model/account inspection, and onboarding/browser flows that are not properly described. Undeclared capabilities are dangerous because they widen the attack surface and can surprise users or calling agents with data access, billing, or external interactions beyond the stated e-commerce image-planning purpose.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
95% confidence
Finding

This finding indicates both over-claiming and undeclared capability expansion, including general chat, video generation, model/account inspection, and onboarding/browser flows that are not properly described. Undeclared capabilities are dangerous because they widen the attack surface and can surprise users or calling agents with data access, billing, or external interactions beyond the stated e-commerce image-planning purpose.

Content

No source excerpt is available for this finding.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
95% confidence
Finding

The skill advertises executable commands, environment-variable use, file access, network access, and shell usage, but declares no explicit tool scope or permissions boundary. In an agent environment, this can cause the skill to inherit broader ambient capabilities than intended, increasing the chance of secret exposure, unauthorized network calls, or filesystem modification if invoked automatically.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The description and usage section define activation and expected user requests entirely in Chinese terms such as '电商套图' and '上架图片', with no indication that other languages are supported or that users may opt into this locale. Under the policy, a skill should not impose a specific language or locale unless it is explicitly optional or clearly justified as region-specific.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The default prompt hard-codes Chinese output ('production-ready Chinese workflow') without any user opt-in or documented locale constraint. This can override user language preference, reduce clarity for non-Chinese users reviewing generated commerce content or code, and increase the chance that important safety, legal, or platform-compliance details are misunderstood.

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 implemented as a general-purpose AI CLI covering chat, image, video, model enumeration, upload, and polling, while the declared skill is narrowly scoped to ecommerce listing image-set generation. This scope mismatch increases the reachable attack surface and enables users or downstream agents to invoke unrelated capabilities that bypass policy and business-intent constraints.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The module docstring explicitly frames the file as a general AI capability tool, contradicting the specialized ecommerce-listing-image skill description. This kind of capability/intent mismatch is dangerous because reviewers, orchestrators, and users may rely on manifest scoping that the code does not actually enforce.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

This code file contains natural-language descriptions and CLI prompts entirely in Chinese, including the module docstring and many user-facing messages. Under the policy rule, forcing a specific language without user opt-in is a locale/language policy violation unless the regional constraint is explicitly documented and justified, which is not present here.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

Arbitrary text chat support is unrelated to the stated ecommerce image-set purpose and effectively turns the skill into a broader general AI interface. In agentic environments, this can be abused to route requests outside the intended domain, weakening policy boundaries and enabling prompt-driven misuse using the skill's configured API access.

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 workflows materially exceed the declared image-focused purpose and permit use cases such as promotional or synthetic video generation that the manifest does not authorize. This discrepancy undermines least privilege and increases the chance of abuse, policy evasion, or accidental high-cost operations through the skill's credentials.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
82% confidence
Finding

The file hard-codes the skill name and description in Chinese in both the module docstring and argument parser description, and repeats that label in generated output metadata. This can be a natural-language locale policy issue because the skill presents a fixed language choice with no documented opt-in or alternative for users in other locales.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Low
Category
Not specified by scanner
Confidence
90% confidence
Finding

Exposing user-info and wallet balance inspection is unnecessary for an image-listing skill and reveals account metadata outside the minimum required functionality. While not a direct compromise by itself, it broadens data exposure and can aid reconnaissance about account state, quotas, or billing.

Content

No source excerpt is available for this finding.