The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill performs runtime compilation of C source code and uses process injection techniques.
scripts/office/soffice.py contains a hardcoded C source string (_SHIM_SOURCE) that it writes to ~/.cache/xlsx-skill/lo-shim/lo_socket_shim.c and compiles using gcc into a shared object.
The resulting library is then injected into the soffice (LibreOffice) process via the LD_PRELOAD environment variable.
While the stated purpose is to work around sandbox restrictions on Unix domain sockets, this technique involves low-level system call interception (socket, listen, accept, close, read).
[COMMAND_EXECUTION]: The skill executes several shell commands to manage its environment and dependencies.
scripts/office/soffice.py executes gcc to compile the shim.
scripts/recalc.py and scripts/office/soffice.py execute soffice (LibreOffice) to perform spreadsheet calculations.
scripts/office/validators/redlining.py executes git diff to perform document comparisons.
[INDIRECT_PROMPT_INJECTION]: The skill represents a potential vulnerability surface for indirect prompt injection via untrusted spreadsheet data.
Ingestion points: The skill reads content from .xlsx, .xlsm, .csv, and .tsv files using pandas.read_excel and openpyxl.load_workbook (referenced in SKILL.md).
Boundary markers: None detected in instructions for handling untrusted workbook content.
Capability inventory: The skill has significant capabilities including arbitrary shell command execution (Bash tool), file writing (Write tool), and the ability to compile and inject code.
Sanitization: While the SKILL.md suggests installing defusedxml to harden against XML expansion attacks, it is not strictly enforced in the provided logic.

xlsx