The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill features dynamic code generation and execution. Specifically, scripts/office/soffice.py contains embedded C source code that is written to temporary storage, compiled at runtime into a shared object (.so) via gcc, and then loaded into the LibreOffice (soffice) process using the LD_PRELOAD environment variable. This is intended to bypass AF_UNIX socket restrictions in restricted environments.
[COMMAND_EXECUTION]: Multiple components execute external system commands using subprocess.run: scripts/recalc.py and scripts/office/soffice.py run the soffice binary for formula processing; scripts/office/soffice.py runs gcc for compilation; and scripts/office/validators/redlining.py runs git diff to compare document revisions during validation.
[PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection. It ingests untrusted data from Excel and CSV files (as described in SKILL.md and implemented in Python processing scripts) and has access to powerful system tools like compiler execution and process injection. The current implementation lacks explicit boundary markers or sanitization logic for the content read from these external files.

xlsx