The Agent Skills Directory

[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection. It is designed to ingest and process data from external sources such as CSV files, Excel workbooks, and Google Sheets, which may contain malicious instructions designed to influence the agent's behavior.
Ingestion points: Data enters via pd.read_csv() in csv.md, load_workbook() in excel.md, and service.spreadsheets().values().get() in google-sheets.md.
Boundary markers: None detected; the skill lacks explicit instructions to ignore embedded commands within the tabular data it processes.
Capability inventory: The skill possesses the ability to write files (to_csv, save), perform network operations (Google Sheets API), and execute complex Python data manipulation logic.
Sanitization: No evidence of data sanitization or validation of the content within the spreadsheets is present.
[EXTERNAL_DOWNLOADS]: The skill relies on several external Python libraries and services for its core functionality.
Dependencies: References the use of pandas, openpyxl, xlsxwriter, and official Google API client libraries (google-oauth2-service-account, googleapiclient.discovery).
Network operations: Performs network requests to Google's Spreadsheet API (spreadsheets.google.com) to read and write data.
[DATA_EXFILTRATION]: While the skill explicitly states it does not store passwords or API keys, it instructs the agent to handle and reference service account keys located at ~/credentials.json. Accessing and referencing these credentials in project metadata files (projects/{name}.md) constitutes the handling of sensitive configuration data.

Spreadsheet