streaming-into-data-lakelisted

<!-- A submission-flavored copy of this skill (with the extra deployment metadata the AWS MCP server registry requires) lives in a separate fork. The version here is the public learning copy. --> # Stream into Data Lake (Firehose to Iceberg) Continuously land rows into an Iceberg table on S3 Tables or a general purpose bucket using Amazon Data Firehose. This skill assumes the destination Iceberg table already exists. Creating the table is owned by `creating-data-lake-table`. The first-time Lake Formation onboarding is owned by `setting-up-lake-formation`. ## Why this skill exists Wiring Firehose to Iceberg through Lake Formation hits four failure modes that no AWS doc surfaces in one place: (1) Firehose silently routes records to the error bucket when the Lambda transformer's output column names do not exactly match the Iceberg schema; (2) Firehose drops Iceberg writes with `AccessDenied` until the Firehose role holds `DESCRIBE` and `SELECT` on the catalog and database plus `SELECT, INSERT, DELETE` on the table; (3) defining the Firehose stream and the Lake Formation grants in one CloudFormation template creates a race because CloudFormation evaluates IAM at stack-create time, before the grants apply; (4) re-deploying the stack with Firehose conditionals enabled but other condition flags omitted resets the unspecified flags to their template default of `false`, silently disabling resources that were already live. This skill encodes the working sequence and the precise er