The Agent Skills Directory

[COMMAND_EXECUTION]: The Python script scripts/agent.py invokes external command-line utilities (vol and pypykatz) using subprocess.run(). This execution uses structured argument arrays and disables shell parsing (shell=True is not used), which limits command injection risks. This behavior is standard and necessary for the skill's primary function of digital forensics and memory analysis.
[INDIRECT_PROMPT_INJECTION]: The skill presents an attack surface for indirect prompt injection due to processing untrusted data sources.
Ingestion points: scripts/agent.py reads raw data from external memory dumps using the windows.strings plugin.
Boundary markers: Missing; the extracted strings are processed directly without delimiters or instruction isolation filters.
Capability inventory: The script includes code execution capabilities via subprocess.run() to execute vol and pypykatz binaries.
Sanitization: Extracted credential tokens are truncated to safe lengths (match[:30]), but the textual data is not sanitized or filtered against prospective LLM injection instructions.

extracting-credentials-from-memory-dump