The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The script scripts/office/soffice.py dynamically generates C source code at runtime, compiles it into a shared object using gcc, and injects it into the LibreOffice process via the LD_PRELOAD environment variable.
[COMMAND_EXECUTION]: The script scripts/recalc.py modifies the LibreOffice user configuration directory by writing a custom StarBasic macro (Module1.xba), allowing for automated code execution within the office environment.
[COMMAND_EXECUTION]: Multiple scripts, including scripts/recalc.py, scripts/office/soffice.py, and scripts/office/validators/redlining.py, use subprocess.run to execute external binaries such as soffice, gcc, and git.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it ingests untrusted spreadsheet data through pandas and openpyxl while maintaining access to shell execution tools. Evidence: Ingestion points are located in SKILL.md and scripts/recalc.py; Boundary markers for external data are absent; Capability inventory includes multiple shell execution points via subprocess.run; Sanitization is limited to defusedxml for XML-based spreadsheet components.

alterlab-xlsx