The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external Google Spreadsheets which could potentially contain malicious instructions.
Ingestion points: Data is pulled into the agent's context through actions like googlesheets_values_get, googlesheets_batch_get, and googlesheets_lookup_spreadsheet_row (SKILL.md).
Boundary markers: The instructions state "Reads first; confirm with the user before any write," which provides a human-in-the-loop checkpoint.
Capability inventory: The skill possesses significant write and management capabilities, including googlesheets_values_update, googlesheets_delete_sheet, and googlesheets_create_chart via the CLI.
Sanitization: No specific technical sanitization or input filtering is described in the prompt; it relies on the agent's internal safety filters.
[EXTERNAL_DOWNLOADS]: The skill relies on npx to fetch and execute the @useclawlink/cli package from the NPM registry. This is expected behavior for this vendor-provided integration tool.
[COMMAND_EXECUTION]: The skill functions by directing the AI agent to execute shell commands through the ClawLink CLI to perform spreadsheet operations and manage authentication.

google-sheets