The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes a local binary lark-cli to perform operations on Lark spreadsheets. The included Python helper scripts (e.g., scripts/lark_sheet_read_cli.py) execute this binary using the subprocess.run function. The implementation correctly passes arguments as a list and avoids the use of shell=True, which is a secure practice to prevent command injection.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to read and process data from external spreadsheets, which represents an attack surface for indirect prompt injection.
Ingestion points: Untrusted data enters the agent's context through tools like +csv-get, +table-get, and +cells-get (documented in references/lark-sheets-read-data.md), as well as via the scripts/lark_profile_table.py script.
Boundary markers: The instructions guide the agent on how to interpret table structures using annotated_csv with row prefixes, but they do not provide explicit delimiters or "ignore instructions" warnings for the data being read.
Capability inventory: The skill possesses significant capabilities, including reading and writing spreadsheet data, creating pivot tables and charts, and performing administrative tasks on the workbook structure via lark-cli calls in scripts/lark_sheet_read_cli.py.
Sanitization: The skill relies on standard data parsing (CSV, JSON) and does not perform additional sanitization or filtering of cell content to detect or neutralize embedded instructions.

lark-sheets