The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8) due to its data-handling design. It is designed to ingest and reason about external configuration files and SQL reporting specs found in the workspace.
Ingestion points: The skill uses Read, Glob, and Grep tools to intake content from arbitrary workspace files (e.g., repository files tied to dashboard delivery).
Boundary markers: The instructions do not define clear delimiters or "ignore embedded instructions" blocks when the agent is reading file content.
Capability inventory: The skill possesses Write and Edit capabilities, allowing the agent to modify the workspace based on instructions that could be maliciously embedded in the analyzed data files.
Sanitization: No validation or sanitization routines are specified for content retrieved via the file-system tools.

looker-studio-bigquery