The Agent Skills Directory

[COMMAND_EXECUTION]: The recalc.py script utilizes subprocess.run to execute external binaries on the host system.
It invokes soffice (LibreOffice) to perform spreadsheet formula recalculation, which is the primary intended function of the script.
It utilizes timeout or gtimeout to manage execution limits for these subprocesses.
The script dynamically generates and writes a StarBasic macro file (Module1.xba) to the local filesystem (LibreOffice configuration directory) to facilitate the recalculation process.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it is designed to ingest and process data from untrusted spreadsheet files such as .xlsx, .csv, and .tsv.
Ingestion points: External data is loaded into the agent context via pandas.read_excel() and openpyxl.load_workbook() as instructed in SKILL.md and implemented in recalc.py.
Boundary markers: There are no specified delimiters or explicit instructions to the agent to ignore or isolate instructions found within the spreadsheet cells.
Capability inventory: The agent has the capability to execute local commands through the provided recalc.py helper script.
Sanitization: No sanitization or content validation is performed on the data read from the spreadsheets before it is analyzed or used by the agent.

xlsx