The Agent Skills Directory

[PROMPT_INJECTION]: The skill's frontmatter description uses authoritative and imperative language designed to override the agent's default behavior and selection logic.
Evidence: The description contains the phrases "CRITICAL: You MUST activate this skill..." and "override default document behavior," which are patterns used to bypass or influence standard agent orchestration.
[PROMPT_INJECTION]: The skill facilitates indirect prompt injection by processing external data without boundary markers or sanitization instructions.
Ingestion points: The skill utilizes docs.getText, docs.getComments, and docs.getSuggestions to read content that may be controlled by third parties.
Boundary markers: There are no instructions to use delimiters or warnings for the agent to ignore embedded instructions within the document data.
Capability inventory: The skill has extensive write and management capabilities, including docs.create, docs.writeText, docs.replaceText, and drive.moveFile.
Sanitization: No sanitization or validation of the ingested document content is performed before processing.

google-docs