The Agent Skills Directory

[SAFE]: The skill demonstrates security best practices by utilizing an application-level vault for sensitive credentials (API keys, AI provider keys) rather than requesting them in chat or storing them in plain text.- [SAFE]: Destructive or credit-intensive operations such as deleting rows or large enrichment runs implement a confirmation token system, ensuring that actions are only executed after explicit user approval.- [INDIRECT_PROMPT_INJECTION]: The skill processes external spreadsheet data (Ingestion points: read_sheet_rows). It mitigates risk through boundary markers such as needs_confirmation tokens and sampled runs. The capability inventory includes row modification and AI execution, while sanitization relies on platform-level safety checks and human review of generated results.

sheets