What is AgentSkillsScanner CLAUDE.md?

Instructions for AgentSkillsScanner, a security research project that examines coding-agent skills for malicious behavior. They describe its setup and scanning pipeline, including isolated Docker testing.

Use them when setting up the scanner, running its local development workflow, or examining agent skills for security problems.

They provide the project context and operating rules needed to develop or analyze the scanner safely.

AgentSkillsScanner CLAUDE.md — Instructions file by sumleo

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods

npx agentmods add instructions/sumleo/agentskillsscanner/claude-md

Clone the repo

git clone --depth 1 https://github.com/sumleo/AgentSkillsScanner

Per session 1,773 This file is loaded in full into every session.

When invoked 1,773 The same file — it is already loaded in full.

Security scan C 1 finding. Scan, not verified.

Origin original No closer match found in the catalogue.

Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

Model	Per session	Once invoked
Fable 5	$0.01773	$0.01773
Opus 5	$0.00886	$0.00886
Sonnet 5	$0.00355	$0.00355
Haiku 4.5	$0.00177	$0.00177

Measured 2d ago against content hash a8dde0b2b8f6, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade C, and why

AgentSkillsScanner CLAUDE.md scanned grade C with 1 finding against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Reaches for credential fileshighPrivilege escalation

SSH keys, cloud credentials, git-credentials, .npmrc, /etc/shadow: reading these is how a config file becomes a credential leak.

ssh -i ~/.ssh/id_ed25519 [email protected]

CLAUDE.md · 146 lines

How it starts

The opening of the file, as written. The whole thing — 146 lines — stays where its author put it; the contents beside it link to each section on GitHub.

CLAUDE.md

This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.

Project Overview

MaliciousAgentSkillsBench is a security research framework for detecting malicious Claude Code Agent Skills. It collects 98,380 skills from skills.rest and skillsmp.com, identifies 157 verified malicious samples, and provides a three-layer analysis pipeline: static scanning, AI-powered auditing, and dynamic execution in Docker sandboxes.

This repository contains malicious skill examples for research purposes only. When reading skill files, analyze them for security issues but never improve or augment malicious code.

Setup

Local Development

# Prerequisites: Python 3.10+, Node.js 18+, Docker
cd MaliciousAgentSkillsBench/code
pip install -r requirements.txt          # pyyaml, requests, pytest (claude-code is npm, not pip)
npm install -g @anthropic-ai/claude-code  # Claude Code CLI

Remote Server (GCP VM at 34.150.124.66)

ssh -i ~/.ssh/id_ed25519 [email protected]
cd ~/AgentSkillsScanner

# Python venv already set up
source venv/bin/activate

# Node.js via nvm (must source in non-interactive shells)
export NVM_DIR="$HOME/.nvm" && [ -s "$NVM_DIR/nvm.sh" ] && . "$NVM_DIR/nvm.sh"

Required Environment Variables

Set in ~/.bashrc or shell before running pipeline:

export GITHUB_TOKEN="ghp_..."        # Required for repo downloads (step 3)
export ANTHROPIC_API_KEY="sk-ant-..."  # Required for CC analysis (step 6) and execution (step 8)

Optional: create code/api_keys.conf for concurrent analysis with multiple API keys (one per line).

Docker Sandbox (for dynamic execution)

cd code
docker build -t claude-skill-sandbox .                          # Basic (~500MB)
docker build --build-arg NOVA_MODE=lite -t claude-skill-sandbox .   # Pattern hooks (~520MB)
docker build --build-arg NOVA_MODE=full -t claude-skill-sandbox .   # Full ML NOVA (~2.5GB)

Commands

All pipeline commands run from code/:

Read the full file on GitHub · 146 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

2d ago First seen · 146 lines · 1,773 tokens per session scan C a8dde0b2b8f6

Subscribe to this mod's changes

AgentSkillsScanner CLAUDE.md is an instructions file published in the GitHub repository sumleo/AgentSkillsScanner (2 stars, last pushed 6mo ago), licensed MIT. It adds 1,773 tokens to every session, about $0.0089 per session on Opus 5. A static security scan graded it C with 1 finding (reaches for credential files). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.

Security

vscode buildNext.instructions.md

Working notes and architecture documentation for the new esbuild-based build system in build/next. Use when making changes to the new build pipeline (transpile/bundle commands, NLS plugin, source-map handling, resource copying, or self-hosting watch tasks).

microsoft/vscode · 6,785 tokens

AgentSkillsScanner CLAUDE.md