The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The script scripts/office/soffice.py dynamically generates C source code, compiles it using gcc, and then uses LD_PRELOAD to inject the resulting shared library into the soffice process. This is a highly aggressive technique used to intercept and modify system calls.
[REMOTE_CODE_EXECUTION]: The script scripts/recalc.py writes a custom LibreOffice Basic macro to the user's local configuration directory (~/.config/libreoffice or ~/Library/Application Support/LibreOffice) and executes it via soffice. This modification persists across application sessions.
[COMMAND_EXECUTION]: Multiple scripts execute system commands using the subprocess module, including gcc for compilation, git for diffing, and soffice for spreadsheet operations. This represents a broad attack surface for command-related exploits.
[PROMPT_INJECTION]: The skill ingests untrusted data from spreadsheets and possesses dangerous capabilities, creating a surface for indirect prompt injection attacks.
Ingestion points: scripts/recalc.py (via pandas and openpyxl).
Boundary markers: Absent.
Capability inventory: Shell command execution via subprocess.run in scripts/recalc.py and scripts/office/soffice.py.
Sanitization: Absent.

xlsx