Security audit

gtrk 封面工作台

Security checks for vulnerabilities and agentic risk

Overview

This cover-design skill also requires extra software and includes video-project editing instructions that contradict its stated scope.

Ask for a cover-only version that removes the unrelated editing instructions and unused CLI dependency. Cover generation does not need a cloud API key. The browser workbench loads external fonts and JavaScript; those dependencies, the CLI implementation, and missing reference documents remain unverified. Empty static-scan results do not resolve the explicit scope contradiction.

A.I.G

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

SkillSpector

Vulnerability Patterns

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
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access

Findings (7)

Description-Behavior Mismatch

High

Category: MCP Tool Poisoning
Confidence: 100% confidence
Finding: 代码中的 gtrk patch move 和 gtrk patch trim 操作视频剪辑项目，不属于封面诊断、Prompt 生成或 H5 排字导出流程，且直接违背描述中“无 gtrk 命令”的明确声明。即使命令使用占位参数，它们所表达的能力仍与声明用途实质不符。
Content: No source excerpt is available for this finding.

Incomplete referenced artifact analysis

High

Category: analysis-evasion
Confidence: 100% confidence
Finding: Referenced artifact was not completely inspected
Content: Scanner excerpt · SKILL.md (reported line 81)May include surrounding context.
md
**Step 4 · 收齐背景图后实例化工作台**（`references/workbench-guide.md` + `assets/workbench-template.html`）：

The workflow mandates English-primary and Chinese-secondary prompts without offering a language choice.

Medium

Category: Output Handling
Confidence: 95% confidence
Finding: Step 2 explicitly requires an English main version and a Chinese supplementary version for every image size, regardless of the user's language preference. Providing both versions is not an opt-in language choice, and the file does not justify this fixed language requirement.
Content: No source excerpt is available for this finding.

The skill directs video-project mutations despite declaring a purely creative workflow with no gtrk commands.

Medium

Category: Excessive Agency
Confidence: 98% confidence
Finding: The manifest limits this skill to cover-design prompts and a browser-based layout workbench, explicitly stating that it has no gtrk commands. However, L114–L125 instruct the agent to use gtrk patch to move and trim video clips and adjust other timeline properties, extending the workflow into modification of the underlying video project rather than cover creation.
Content: No source excerpt is available for this finding.

Requires English-primary and Chinese-secondary output without user opt-in.

Medium

Category: Output Handling
Confidence: 96% confidence
Finding: The instruction mandates always producing both an English main version and a Chinese supplementary version, regardless of the user's language preference. Although later sections discuss model-specific language suitability, they do not justify requiring both languages for every output.
Content: No source excerpt is available for this finding.

The T5 decorative-text slot requires English without user opt-in.

Low

Category: Output Handling
Confidence: 90% confidence
Finding: The scaffold specifies English for the generated T5 text, reinforcing the English-only slot definition at L010. It does not offer a user language choice or explain why this text must be English.
Content: No source excerpt is available for this finding.

The template instructions require English subtitles and decorative text without user opt-in.

Low

Category: Output Handling
Confidence: 92% confidence
Finding: L014 specifies an English subtitle (英文副题), and L018 specifies decorative English text (装饰英文). These requirements prescribe a language for generated cover content without offering a user choice or documenting a locale-specific justification.
Content: No source excerpt is available for this finding.