The Agent Skills Directory

Indirect Prompt Injection Surface: The skill is designed to ingest and process information from external sources, specifically Google Docs content and Google Calendar event data. This capability introduces a potential surface for indirect prompt injection, where data from these sources might contain instructions intended to deviate the agent from its primary task. However, the skill incorporates significant mitigations, such as strict "target-document guards," mandatory connector-readback verification after writes, and a requirement to resolve structural anchors from live data before any modification, which reduces the risk of the agent following unintended instructions.
Connector-Based Security Model: The instructions explicitly prioritize using native Google Docs and Drive connectors for all document operations. It specifically discourages the use of local helper scripts or code-mode bridge writes for document generation. This focus on standard, managed API interactions rather than arbitrary code execution on the local system significantly narrows the potential attack surface for common local execution vulnerabilities.
Vendor-Managed Dependencies: The skill utilizes the google-drive@openai-curated plugin and the documents plugin. These resources originate from the skill's author (OpenAI), representing a standard integration of vendor-owned tools within their specific runtime environment. The use of these managed plugins is consistent with the skill's intended purpose of document management and authoring.

google-docs