The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses subprocess.run in scripts/agent.py to execute forensic tools including foremost, scalpel, and file. All subprocess calls use argument lists rather than shell strings, which effectively prevents shell command injection vulnerabilities.
[PRIVILEGE_ESCALATION]: The setup instructions in SKILL.md include sudo apt-get install commands to install the necessary forensic utilities. This use of elevated privileges is standard for software installation on a forensic workstation and is appropriate for the skill's domain.
[DYNAMIC_EXECUTION]: The workflow in SKILL.md includes Python script blocks executed via shell redirection (e.g., python3 << 'PYEOF'). These scripts automate the validation of carved files and the generation of evidence catalogs, representing a routine use of script generation for task automation.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted disk images and search through recovered file contents (e.g., searching for keywords in carved PDFs). While this creates a theoretical attack surface for indirect prompt injection from the processed data, the skill does not explicitly feed the extracted content back into the agent's control loop for instruction following.
Ingestion points: Raw disk images and carved files (PDFs, images, etc.) processed by forensic tools and Python scripts.
Boundary markers: None identified in the prompt interpolation steps.
Capability inventory: File system access (read/write/delete), command execution for forensic tools, and logging.
Sanitization: No specific content filtering is applied to the data extracted from the disk images before metadata extraction or keyword searching.

performing-file-carving-with-foremost