Introduction

With the increasing number of Skills in the community, directories often come with scripts, networking descriptions, and long lists of promises in SKILL.md. It is difficult to determine by visual inspection alone whether they secretly read local files, make external requests, or if there is a discrepancy between what the documentation claims and what the implementation actually does. The following introduces the “Skill Safety Review” on SkillHub, helping you clarify common risk points before officially enabling a package.

What is this

“Skill Safety Review” (ID @user_909957f0/skill-safety-check, Version 1.0.0) is a Skill from the SkillHub community (community), categorized under it-ops-security, and maintained by the publisher user_909957f0.

Its official positioning is to perform a more granular security review based on the skill-vetter concept. It targets risks such as scripts, networking behavior, local file access, keys, and hidden instructions, allowing you to judge whether a Skill is trustworthy before installation. The page tag indicates requires_api_key=false, meaning using this Skill itself does not require an additional API Key.

Core Functions

Combined with the SkillHub page introduction, the verified capability boundaries are as follows:

  1. Pre-installation Audit: Perform a security-oriented review of the skill directory before actually installing a Skill (both community packages and self-developed packages can be reviewed).
  2. Scripts: Analyze high-risk clues in the scripts and related implementations within the package.
  3. Networking: Check for descriptions or implementation traces of external networking or calls to external services.
  4. Local Files: Focus on risks related to the access, reading, and writing of local files.
  5. Keys: Identify issues regarding the exposure of sensitive information such as keys and tokens.
  6. Hidden Instructions: Identify hidden instruction risks that may mislead model behavior (an extension of the skill-vetter concept).

This Skill is indicated to perform more detailed analysis on top of skill-vetter; specific inspection steps are subject to the SKILL.md within the package after installation.

Installation and Activation

The installation method provided on the SkillHub page involves sending a prompt to your AI assistant, asking it to fetch the Skill according to the SkillHub installation documentation:

Please install `@user_909957f0/skill-safety-check` based on https://skillhub.cn/install/skillhub.md.

For different clients (Cursor, Codex CLI, Claude Code, etc.), the final directory and activation method are subject to the instructions in https://skillhub.cn/install/skillhub.md. This article does not detail client-specific differences to avoid confusion with unverified configuration items.

Typical Usage

  1. Obtain a Skill archive or local directory from SkillHub or others; do not install it directly into the production environment immediately.
  2. In the conversation, specify that you want to perform a security review on a “specific Skill folder” and point to the directory path or the extracted root directory (the package root containing SKILL.md).
  3. Ask the assistant to follow the “Skill Safety Review” process to provide conclusions and evidence regarding dimensions such as scripts, networking, local files, keys, and hidden instructions.
  4. Decide whether to install, share, or request modifications from the author based on the review results.

The official documentation does not provide a fixed one-click CLI command; the typical path is to enable the Skill and then trigger the review task via natural language.

Applicable Scenarios and Notes

Who it is for

  • Developers or teams who frequently install third-party Skills from the community.
  • Those who need to perform a reproducible security self-check before deploying to an intranet or client environment.
  • Those who want to perform a local self-check before packaging a Skill for external distribution.

Notes

  • Review conclusions for Community Skills (community) depend on the current package content and the execution quality of the assistant; this tool is not a substitute for professional penetration testing or compliance audits.
  • SkillHub page statistics (as of data capture): Rating 100000, Downloads 1536, Installations 0; popularity and download counts will change, subject to the SkillHub Skill Page.
  • If scripts within the package are to be read or statically analyzed during the review process, it is recommended to operate on an isolated copy to avoid accidentally executing unknown commands.

Conclusion

“Skill Safety Review” turns the idea of “taking a look before installing” into a repeatable checklist: an extension based on skill-vetter, focusing on risks such as scripts, networking, local files, keys, and hidden instructions, and helping to compare against documentation claims. If needed, you can install @user_909957f0/skill-safety-check on SkillHub, or open the official entry: https://api.skillhub.cn/user_909957f0/skill-safety-check (see SkillHub Skill Page for page navigation).