The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill presents an attack surface where malicious instructions embedded in processed files could influence the agent's behavior. Ingestion points: The skill can ingest untrusted data from the local file system using the Read tool. Boundary markers: The instructions do not include specific delimiters or 'ignore embedded instructions' warnings to prevent the agent from following commands found within data it processes. Capability inventory: The skill is authorized to perform impactful operations using Write, Edit, and Bash(gcloud:*) tools, which could be misused if an injection occurs. Sanitization: There are no requirements defined for the agent to sanitize or validate external content before using it in configuration generation or command execution.

cloud-logging-sink-setup