The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides multiple scripts and operational recipes that execute the ftkimager.exe command-line utility to perform bitstream acquisitions of physical drives and volatile memory.
[PRIVILEGE_ESCALATION]: The instructions explicitly direct the user to run forensic acquisition tools as an elevated Administrator to bypass operating system restrictions on physical device access. Evidence is found in SKILL.md ("Execute ftkimager.exe strictly as elevated Administrator") and gpt.md ("Launch PowerShell/CMD as elevated Administrator").
[INDIRECT_PROMPT_INJECTION]: The skill establishes an attack surface by ingesting and processing untrusted external evidence data.
Ingestion points: The harvest_attachments function in gemini.md recursively parses .eml files from external directories.
Boundary markers: None are present in the provided scripts to delimit extracted content from agent instructions.
Capability inventory: The skill has capabilities for command execution (ftkimager.exe) and file-system modifications across all core scripts.
Sanitization: The scripts use standard libraries to extract data but do not perform validation or sanitization of the content before processing or logging.
[CREDENTIALS_UNSAFE]: The database mining script ftk_db_miner.py in gpt.md includes a hardcoded default password (adpassword) for the PostgreSQL backend. While this is the well-known default for the AccessData FTK installation, it represents a hardcoded credential in a script template.

ftk