What this skill does
databricks-bundle-medic is the deploy and infrastructure skill of the pack. It covers the Databricks Asset Bundles surface (which is immature tooling — the replacement for the deprecated dbx, with a moving bug list across CLI versions) plus the deploy-time infrastructure operations that bite production teams when they touch encryption keys, networking topology, or workspace-level configuration that compute interacts with at boot.
What it catches
Design questions I want pushback on
- PreToolUse hook on
databricks bundle deploy. Plan is: download the remote terraform.tfstate, validate it parses as JSON, warn if size has shrunk vs. last known good copy, cache a known-good copy locally as a recovery escape hatch. Right behavior, or too cautious?
- PostToolUse auto-retry on D6. On the specific "User does not have CREATE TABLE on Schema" stderr pattern, auto-retry once. Plan is to fire ONLY on this exact known-transient failure, never on others. Right call, or risky?
databricks bundle bind workaround scope. D4 — should the skill ship a terraform.tfstate editor (powerful but dangerous), or only refuse to proceed and direct the user to wait for upstream fix (safe but unhelpful)?
- CMK rotation runbook. D8 needs a workspace-drain sequence. Plan is to enumerate all clusters/pools/warehouses, ask for confirmation, terminate in dependency order, run rotation, restart. Should the skill do the termination, or only produce the sequence as a runbook the user runs manually?
- PrivateLink audit detection. D9 is the easiest to detect (enumerate VPC endpoints, check for S3/STS/Kinesis coverage) but the report needs cost math. Are the AWS cost-explorer integrations worth the auth complexity, or is "here are the missing endpoints, expected NAT cost is X" enough?
What I am not asking about right now
- Whether to support
dbx migration — out of scope, dbx is deprecated.
- Whether to handle Azure-specific networking — Azure ExpressRoute and PrivateLink Service work is acknowledged but deferred to v3.
- Whether to add Terraform state import features — Terraform-side problems live in the Terraform provider's issue tracker.
How to respond
Comment below with any thoughts, leave thumbs-up / thumbs-down on individual bullets in the design questions, or send a voice memo on WhatsApp and I will transcribe it into the issue with attribution. English is not required for voice memos — Portuguese is fine.
Source bead: claude-jhnj in the local beads workspace.
Reference material
Most relevant for this skill:
| Doc |
What it covers |
004-RL-RSRC |
Unity Catalog / Asset Bundles / identity / workspace ops / secrets / networking pain catalog |
Full reference set + cross-skill context: see umbrella issue #795 § Reference material.
- Jeremy Longshore
intentsolutions.io
What this skill does
databricks-bundle-medicis the deploy and infrastructure skill of the pack. It covers the Databricks Asset Bundles surface (which is immature tooling — the replacement for the deprecateddbx, with a moving bug list across CLI versions) plus the deploy-time infrastructure operations that bite production teams when they touch encryption keys, networking topology, or workspace-level configuration that compute interacts with at boot.What it catches
databricks bundle bindmissing UC catalog and external location support — teams with existing UC resources cannot bring them under DAB management without editingterraform.tfstateby hand or destroying and recreating the resource. Open at Bundle bind does not support Unity Catalog resources (catalogs, external locations) databricks/cli#4842 as of CLI v0.295.x. (004-RL-RSRCD4)DATABRICKS_BUNDLE_ENGINE=direct(officially preview, in practice adopted permanently). Open at bundle deploy fails with "unexpected EOF" reading terraform.tfstate on every deploy after initial databricks/cli#4986. (004-RL-RSRCD5)004-RL-RSRCD6)004-RL-RSRCD8)004-RL-RSRCD9)Design questions I want pushback on
databricks bundle deploy. Plan is: download the remoteterraform.tfstate, validate it parses as JSON, warn if size has shrunk vs. last known good copy, cache a known-good copy locally as a recovery escape hatch. Right behavior, or too cautious?databricks bundle bindworkaround scope. D4 — should the skill ship aterraform.tfstateeditor (powerful but dangerous), or only refuse to proceed and direct the user to wait for upstream fix (safe but unhelpful)?What I am not asking about right now
dbxmigration — out of scope, dbx is deprecated.How to respond
Comment below with any thoughts, leave thumbs-up / thumbs-down on individual bullets in the design questions, or send a voice memo on WhatsApp and I will transcribe it into the issue with attribution. English is not required for voice memos — Portuguese is fine.
Source bead:
claude-jhnjin the local beads workspace.Reference material
Most relevant for this skill:
004-RL-RSRCFull reference set + cross-skill context: see umbrella issue #795 § Reference material.
intentsolutions.io