The Agent Skills Directory

[COMMAND_EXECUTION]: The skill's core functionality relies on executing external security tools. The scripts scripts/code_security_scanner.py, scripts/dependency_auditor.py, and scripts/security_scanner.py use subprocess.run() with list-style arguments (shell=False) to safely invoke system utilities like bandit, npm audit, and pip-audit. This avoids shell injection risks.
[EXTERNAL_DOWNLOADS]: The scripts/dependency_auditor.py script includes logic to dynamically install the pip-audit package from the official Python Package Index (PyPI) if it is missing from the environment. This targets a well-known service and is consistent with the skill's purpose of managing security dependencies.
[SAFE]: The skill demonstrates excellent defense-in-depth by explicitly listing high-risk utilities (e.g., nmap, sqlmap, curl, rm) in the disallowed-tools section of the SKILL.md frontmatter. This prevents the agent from being misused as an arbitrary attack tool.
[PROMPT_INJECTION]: The skill processes untrusted data from scanned source code and HTTP responses. While this creates a surface for indirect prompt injection, the orchestrator instructions and scripts are focused on structured vulnerability reporting rather than command execution based on scanned content. A default 'governance-first' workflow is enforced to ensure all activities are authorized.
[DATA_EXFILTRATION]: Network operations are restricted to HTTP scanning of user-defined target URLs. No evidence was found of sensitive local files (like credentials or environment variables) being transmitted to unauthorized external domains.

performing-penetration-testing