ClawHub
Back to skill

Security audit

Flexible Database De

Security checks for vulnerabilities and agentic risk

Overview

This skill gives SQLite database-design guidance and asks agents to create local project files and run validation commands, which is disclosed and aligned with its purpose but should be used with normal project-write caution.

Install this only if you want an agent to help create or modify files in a local SQLite project. Ask the agent to show the exact files and commands before it runs them, especially any package install or script execution, and be aware that some referenced scripts appear to be described but not included in this artifact.

SkillSpector

Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Behavioral ASTexec() Call, eval() Call, Dynamic Import
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Findings (3)

Vague Triggers

Medium
Confidence
86% confidence
Finding
The trigger keywords are overly broad and generic, including common words like 'knowledge', 'users', 'guide', 'implement', 'sqlite', and 'design'. This can cause unintended activation of the skill in unrelated conversations, increasing the chance that the agent will propose or perform file creation and shell-backed setup actions in contexts the user did not explicitly request.

Missing User Warnings

Medium
Confidence
95% confidence
Finding
The skill instructs the agent to copy scripts into a user project, create directories, and run Python commands, but it does not require an explicit user-facing warning or confirmation before modifying the filesystem or executing commands. In an agent environment with exec access, unintended invocation or over-eager automation could lead to unauthorized local changes, execution of unreviewed scripts, or disruption of a user's workspace.

Unrestricted Tool Access

Medium
Category
Excessive Agency
Content
## 工具声明（Skill平台 / SkillHub）

allowed-tools:

* Bash
* FileRead
Confidence
90% confidence
Finding
The skill declares broad tool access including Bash/Shell plus file read/write capabilities, while also containing workflow steps that direct project modification and command execution. In the context of a markdown-driven agent skill, this combination increases the blast radius of accidental or unintended activation, because the skill can move from giving advice to directly changing the local environment.

Static analysis

No suspicious patterns detected.