The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill establishes a federation to external, third-party data catalogs (Databricks Unity and AWS Glue). Instructions explicitly direct the agent to 'Summarize to the user what kind of data was found' and 'Propose a business or analytical question', creating a pathway where instructions embedded in external data could influence agent behavior.
Ingestion points: External catalog metadata and table contents via bq ls and SELECT queries.
Boundary markers: None. The skill does not instruct the agent to ignore instructions embedded in the retrieved data.
Capability inventory: Execution of gcloud, aws, and bq commands; local file creation (credentials.json, trust_policy.json).
Sanitization: No validation or sanitization of external data is specified before processing.
[COMMAND_EXECUTION]: The skill requires the execution of multiple high-privilege administrative commands using the Google Cloud CLI (gcloud), AWS CLI (aws), and BigQuery CLI (bq) to manage IAM policies, secrets, and cloud resources.
[CREDENTIALS_UNSAFE]: The procedure instructs the user to create a local JSON file (credentials.json) containing a client_id and client_secret. While intended for Secret Manager upload, local storage of plaintext credentials increases the risk of exposure if the environment is compromised or the file is not subsequently deleted.

federate-lakehouse-catalog