The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses python3 to execute local scripts (read_tracking_excel.py, check_property_duplicates.py, generate_output_csv.py) for processing tracking data. These scripts are included within the skill's package and perform standard JSON and Excel parsing tasks.
[EXTERNAL_DOWNLOADS]: The skill requires the openpyxl Python library for Excel processing. This is a standard, well-known package from the Python Package Index (PyPI) used for spreadsheet manipulation.
[DYNAMIC_CONTEXT_INJECTION]: The SKILL.md contains examples of dynamic tool usage using lark-cli, a command-line tool for Feishu/Lark integration. The usage is restricted to authenticated spreadsheet operations (sheets +cells-clear, +csv-put, +batch-update) and is triggered only after explicit user confirmation in Step 3.5.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from Excel workbooks and strategy texts (Step 1 and Step 2). It employs boundary markers (Workflow and Validation Gates) and explicit user review steps before any automated writing occurs to prevent unintended actions based on the ingested content.
[DATA_EXFILTRATION]: While the skill can write to external Feishu spreadsheets, this is the primary declared purpose of the skill. It uses authenticated lark-cli commands and requires the user to supply the destination URL and confirm the scope. No unauthorized or silent exfiltration was detected.

push-tracking-gap-analyzer