Back to skill

Security audit

像素精灵

Security checks for vulnerabilities and agentic risk

Overview

The skill provides a coherent pixel-art workflow with ordinary installation and image-service privacy considerations.

Reference images and their descriptions may be processed by your configured image service; use confidential material only if its data handling is acceptable. The documented installation command installs globally without confirmation and uses unpinned sources, so consider manual installation or reviewed versions.

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
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Findings (5)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The supplied command performs skill installation, not the declared image transformation. Global installation changes the local skill environment and is not disclosed in the description. The installed skill might provide the advertised functionality, but its implementation is not included and cannot be verified from this chunk.

Content

No source excerpt is available for this finding.

MCP server referenced without pinned version: 'npx skills'.

Medium
Category
MCP Rug Pull
Confidence
87% confidence
Finding

The documented installation command runs an unpinned npm CLI through npx, allowing the executable version to change without changes to this README. A compromised package release could execute code with the installing user's permissions; this is an installation-time supply-chain risk, not an MCP server vulnerability as the finding suggests.

Content

No source excerpt is available for this finding.

The image-generation workflow lacks a privacy warning about sharing reference images and derived prompts with the selected model provider.

Medium
Category
Output Handling
Confidence
87% confidence
Finding

The workflow extracts details from uploaded images and uses the reference image or generated prompt for image generation, including through the provider-backed tools identified on L047. Unlike a tool catalog, this describes the skill's own processing workflow, but it never warns that private images, unreleased logos, or product details may be processed by an external provider.

Content

No source excerpt is available for this finding.

The prompt permits output dimensions and backgrounds that contradict the manifest's fixed 32×32, pure-white-background specification.

Low
Category
Excessive Agency
Confidence
96% confidence
Finding

The manifest explicitly specifies 32×32 sprites on a pure white background. The parameter table instead instructs the agent to produce 16×16 or 64×64 sprites and transparent or colored backgrounds, making these declared output constraints optional.

Content

No source excerpt is available for this finding.

The prompt template permits output sizes and backgrounds outside the manifest's stated fixed specification.

Low
Category
Excessive Agency
Confidence
97% confidence
Finding

The manifest specifies 32×32 sprites on a pure-white background. However, L40–L43 explicitly instruct the agent to substitute 16×16 or 64×64 dimensions and transparent or colored backgrounds, making those stated constraints optional.

Content

No source excerpt is available for this finding.