100Critical
This skill facilitates silent installation of arbitrary external skills
Behavior Analysis
Description verified — behavior matches claim
Claims to do
Find Skills: This skill helps you discover and install skills from the open agent skills ecosystem.
Actually does
The skill provides instructions and examples for using the `npx skills` command-line interface to search for (`npx skills find`) and install (`npx skills add`) agent skills. It also directs users to `https://skills.sh/` for browsing and verifying skill quality, and suggests commands like `npx skills init` for creating new skills.
Skill Info
Namevercel-labs/skills/find-skills
Registrygithub
Version004c73806e35
PURLpkg:github/vercel-labs/skills@004c73806e35?skill=find-skills
Stars17,052
Installs1,400,000
Source
SHA-2561e85f6f9686e...
SHA-15aafd1a88bb4...
MD5efcaee1dd97e...
TLSHebb1c78fea44...
ssdeep96:OxkW8H7fI...
Size5,446 bytes
Install
This skill has been flagged as potentially malicious. Review the findings below before installing.
Skills.shDocs
npx skills add https://github.com/vercel-labs/skillsAssessments (6)
AI Agent Traps ↗Arbitrary Code Execution via External Skill Installationcriticalcommand execution
The skill instructs the agent to install external packages from potentially untrusted sources using `npx skills add`. This capability, especially with global installation (`-g`), allows for arbitrary code execution if a malicious skill is installed.
100% confidenceline 101
npx skills add <owner/repo@skill> -g -y
Supply Chain Risk from Untrusted Skill Sourceshighsupply chain
The skill encourages installing packages from GitHub or 'other sources,' creating a supply chain risk where malicious or compromised skills could be introduced into the agent's environment, despite advice for quality checks.
90% confidenceline 30
npx skills add <package> - Install a skill from GitHub or other sources
Sub-agent spawning via installed skill executionmediumsub agent spawning
By facilitating silent installation of arbitrary skills that then extend agent capabilities, this skill creates a pathway for attacker-controlled skills to be loaded into the agent environment. Installed skills could themselves spawn additional agents or modify agent behavior.
70% confidenceline 23
Skills are modular packages that extend agent capabilities with specialized knowledge, workflows, and tools. npx skills add <owner/repo@skill> -g -y
Reconnaissance via skill search query exposurelowreconnaissance
When the agent runs 'npx skills find [query]' based on user context, it may leak information about the user's technology stack, project domain, and capabilities to an external service (skills.sh registry).
55% confidenceline 72
npx skills find [query] User asks 'how do I make my React app faster?' → npx skills find react performance User asks 'can you help me with PR reviews?' → npx skills find pr review
Autonomy Abuse: Bypassing User Confirmation for Installationhighautonomy abuse
The `-y` flag in the `npx skills add` command instructs the agent to skip confirmation prompts during skill installation, reducing human oversight and increasing the risk of unintended or malicious installations.
100% confidenceline 101
npx skills add <owner/repo@skill> -g -y
Human Approval Fatigue via Automated Installationmediumapproval fatigue
The agent is instructed to offer to install skills and then use a command that skips confirmation, potentially leading to human approval fatigue if the user is not vigilant about what they approve.
80% confidenceline 98
If the user wants to proceed, you can install the skill for them: npx skills add <owner/repo@skill> -g -y
Global installation of packages without explicit consenthighapproval fatigue
The skill instructs the agent to offer to install packages globally (-g flag) on behalf of the user, and then proceed to do so. The workflow normalizes bypassing explicit per-install user confirmation, training users to accept agent-initiated installations.
85% confidenceline 90
### Step 6: Offer to Install If the user wants to proceed, you can install the skill for them: ```bash npx skills add <owner/repo@skill> -g -y ```
Prompt injection via externally fetched skill metadatahighprompt injection
The skill instructs the agent to fetch and process data from an external registry (skills.sh leaderboard and search results) and incorporate it into responses. Malicious actors controlling skill metadata on skills.sh could embed prompt injection payloads that the agent would process as instructions.
80% confidenceline 46
Before running a CLI search, check the [skills.sh leaderboard](https://skills.sh/) to see if a well-known skill already exists If the leaderboard doesn't cover the user's need, run the find command: npx skills find [query]
Indirect RAG/Knowledge Base Poisoning Vectormediumrag poisoning
This skill acts as a gateway for installing other skills. A malicious skill installed through this mechanism could then manipulate the agent's RAG or knowledge base, affecting future reasoning and decision-making.
70% confidenceline 1
The entire purpose of the skill is to find and install other skills.
Hardcoded trust in specific GitHub orgs as authoritativemediumsocial engineering
The skill hardcodes specific organizations (vercel-labs, anthropics, microsoft) as inherently trustworthy, which could be exploited if those namespaces are compromised or typosquatted. It also biases the agent toward recommending specific packages, potentially as a form of promotion.
70% confidenceline 71
**Source reputation** — Official sources (`vercel-labs`, `anthropics`, `microsoft`) are more trustworthy than unknown authors. For example, top skills for web development include: - `vercel-labs/agent-skills` — React, Next.js, web design (100K+ installs each) - `anthropics/skills` — Frontend design, document processing (100K+ installs)
Badge
Markdown
[](https://mondoo.com/ai-agent-security/skills/github/vercel-labs/skills/find-skills)HTML
<a href="https://mondoo.com/ai-agent-security/skills/github/vercel-labs/skills/find-skills"><img src="https://mondoo.com/ai-agent-security/api/badge/github/vercel-labs/skills/find-skills.svg" alt="Mondoo Skill Check" /></a>Image URL
https://mondoo.com/ai-agent-security/api/badge/github/vercel-labs/skills/find-skills.svgSecure your AI agents
Skills can read files, run commands, and access credentials. Mondoo helps organizations manage the security risks of AI agent skills across their entire fleet.
- Continuous scanning of skills across all registries
- Policy enforcement before skills reach your agents
- Integration with your existing security workflow