The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/agent.py executes PowerShell commands using subprocess.run to interact with Windows mutexes.\n- [DYNAMIC_EXECUTION]: The agent.py script dynamically builds PowerShell commands via string interpolation and executes them. It also includes a generate-script command that creates a new PowerShell file (mutex_vaccination.ps1) on the filesystem.\n- [INDIRECT_PROMPT_INJECTION]: The skill contains a vulnerability surface where the mutex_name parameter is interpolated into PowerShell command strings without sanitization in the check_mutex_exists_windows and create_mutex_windows functions. A malicious actor could provide a mutex name containing shell metacharacters to execute arbitrary PowerShell commands.\n
Ingestion points: Mutex name strings processed in agent.py.\n
Capability inventory: subprocess.run (shell command execution) and file system write access in agent.py.\n
Sanitization: Absent; the code uses direct f-string interpolation into a shell command context.\n- [METADATA_POISONING]: There is an inconsistency in attribution; the YAML frontmatter in SKILL.md lists the author as 'mahipal', whereas the LICENSE file attributes copyright to 'mukul975'.

implementing-ransomware-kill-switch-detection