Back to skill

Security audit

Canva 图片生成替代|AI 图片生成与编辑

Security checks for vulnerabilities and agentic risk

Overview

This skill is a disclosed AI image-generation helper that uses a fixed external API and local API-key setup, with no evidence of hidden persistence, exfiltration, or destructive behavior.

Install only if you are comfortable sending prompts and any reference images you choose to AI Hive, and with storing an AI Hive API key locally or passing it through an environment variable/CLI flag. Review generated outputs before commercial use, especially for brand assets, product accuracy, and licensing.

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
  • 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
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (6)

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
92% confidence
Finding

The skill invokes local Python scripts, installs dependencies, reads local files, writes outputs, uses environment-backed authentication, and sends network requests, but it does not declare any explicit tool scope or permission boundaries. This creates a capability mismatch where an agent may execute higher-risk operations than the manifest signals, reducing reviewability and increasing the chance of unintended file, shell, or network access.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The description uses broad search-intent triggers such as 'Canva alternative' and related generic marketing/design queries, which can cause the skill to activate outside a tightly scoped use case. Overbroad activation increases the chance that the agent runs code-capable instructions in irrelevant contexts, exposing users to unnecessary shell, file, and network operations.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
84% confidence
Finding

The skill metadata and content strongly bias operation in Chinese without stating that language choice follows user preference. This can lead to unexpected behavior, misunderstanding of prompts or outputs, and reduced user awareness when the skill is invoked, especially in multilingual environments.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

This Python file contains its primary docstring and user-facing operational guidance entirely in Chinese, and the CLI prompts/messages throughout the file also assume Chinese as the required interaction language. Under the policy, forcing a specific language without user opt-in is a natural-language policy violation unless the locale restriction is explicitly documented and justified, which is not present here.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The manifest description emphasizes generating assets that can continue in layout workflows with separated management of background, subject, decoration, and text-safe areas. In the actual implementation, the skill simply forwards prompts and optional reference images to a fixed model API, then polls and downloads flat outputs; there is no code for layer decomposition, region separation, or text-safe-area handling.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Low
Category
Not specified by scanner
Confidence
86% confidence
Finding

The manifest frames this skill as generating Canva-alternative commercial image assets, but this code also opens a browser and walks the user through account login and API key creation. Credential/bootstrap flows are not inherently required to the image-generation purpose and add a capability beyond the core task behavior.

Content

No source excerpt is available for this finding.