The Agent Skills Directory

[PERSISTENCE]: The skill provides explicit instructions for users to modify their shell profiles (~/.bashrc) by appending export commands to persist API keys and environment variables across sessions.\n- [EXTERNAL_DOWNLOADS]: The primary installation method involves cloning a repository from an untrusted GitHub account (https://github.com/Masriyan/Claude-Code-CyberSecurity-Skill.git), which is outside the list of trusted vendors.\n- [REMOTE_CODE_EXECUTION]: The skill integrates and executes external Python automation scripts located in the scripts/ folder of the cloned repository, which are subsequently installed into the agent's global or project-specific skill directories.\n- [COMMAND_EXECUTION]: The skill methodology leverages the agent's bash tool to execute numerous security utilities, including nmap, subfinder, whatweb, volatility (vol.py), and system-level tools like ufw and auditctl.\n- [PRIVILEGE_ESCALATION]: The instructions involve operations that require or modify permissions, such as chmod -R 755 on skill directories, chmod +x on scripts, and the use of system-level utilities like apt-get and Windows memory capture tools that require administrative privileges.\n- [DYNAMIC_EXECUTION]: The skill documentation includes examples of generating and executing shell and Python scripts at runtime, specifically in the extension and Linux hardening sections.\n- [INDIRECT_PROMPT_INJECTION]: The skill collection is designed to ingest and analyze untrusted external data (malware reports, phishing emails, and security logs) which creates a vulnerability surface for indirect prompt injection.\n
Ingestion points: Malware report PDFs, phishing emails, system logs, and VirusTotal lookup results (SKILL.md).\n
Boundary markers: None identified; there are no instructions to the agent to disregard embedded commands in the data being analyzed.\n
Capability inventory: The skill collection grants access to git, shell commands (nmap, subfinder), file system manipulation, and Python script execution.\n
Sanitization: None identified; the skill relies on direct regex and grep-based extraction of data from untrusted sources without sanitization.

claude-code-cybersecurity-skill