The Agent Skills Directory

[Dynamic Execution] (MEDIUM): The skill utilizes python3 -c to execute dynamically constructed Python code. It provides templates with placeholders like <file> and <pattern>. If an agent interpolates untrusted or unescaped strings into these placeholders, it could lead to command injection or arbitrary code execution within the Python sub-process.
[Indirect Prompt Injection] (LOW): The skill is designed to process untrusted external data (disk images and binary files) which could contain malicious payloads targeting the agent's parsing logic.
Ingestion points: Python open().read() and shell strings/grep operations on external files.
Boundary markers: None. The skill does not implement delimiters or 'ignore' instructions for the data it processes.
Capability inventory: File system read access, shell utility execution, and Python interpretation.
Sanitization: None. The skill relies on standard utilities without input validation.
[Data Exposure] (LOW): By design, the skill searches for passwords and sensitive fragments. While this is the primary purpose (forensics), it provides a functional template for identifying sensitive data within accessible files. It does not, however, target specific system-level secret paths like SSH keys or environmental configs.

password-recovery