The Agent Skills Directory

[COMMAND_EXECUTION]: The recalc.py script executes system commands using subprocess.run to interact with LibreOffice (soffice) and system timeout utilities. While necessary for the skill's recalculation feature, this involves programmatic interaction with external system binaries.
[METADATA_POISONING]: There is a discrepancy between the reported author ('spacezephyr') and the copyright notice in LICENSE.txt ('Anthropic, PBC'). This inconsistency in metadata could be misleading regarding the origin and ownership of the skill.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from external Excel, CSV, and TSV files, which serves as an ingestion point for potentially malicious instructions embedded in spreadsheet content.
Ingestion points: The SKILL.md documentation instructs the agent to use pd.read_excel() and load_workbook() to ingest data from external files into the agent's context.
Boundary markers: The instructions do not provide specific delimiters or ignore-instructions for the agent when processing data from spreadsheets, increasing the risk of the agent following instructions found within cells.
Capability inventory: The skill has the capability to read and write local files and execute system commands through the recalc.py script.
Sanitization: There is no evidence of data validation or sanitization logic to filter out natural language instructions from the spreadsheet data being processed.

xlsx