The Agent Skills Directory

[COMMAND_EXECUTION]: The Python scripts scripts/agent.py and scripts/process.py utilize subprocess.run to call system utilities like bcdedit, powershell, and readelf. These commands are used to inspect and report on system configuration and process metadata for security auditing purposes.
[PRIVILEGE_ESCALATION]: Instructions within SKILL.md and the audit scripts involve high-privilege operations, such as modifying the Windows Boot Configuration Data (bcdedit /set nx AlwaysOn) and system-wide exploit protection settings. These activities are appropriate and necessary for the skill's stated objective of endpoint security hardening.
[INDIRECT_PROMPT_INJECTION]: The skill exhibits a theoretical attack surface for indirect injection through metadata processing:
Ingestion points: File names and paths are read via os.walk in scripts/agent.py, and process names are retrieved via PowerShell in scripts/process.py.
Boundary markers: None present; the tools process the collected metadata directly as input for audit utilities.
Capability inventory: The skill executes subprocesses (readelf, powershell) using the collected metadata and writes results to local JSON files.
Sanitization: No explicit sanitization of file paths or process names is performed before they are passed to audit commands, though the use of subprocess.run without shell=True prevents direct shell command injection.

implementing-memory-protection-with-dep-aslr