gke-labs/gke-agentic-migration/bootstrapping
Claude Code and agent skill by gke-labs/gke-agentic-migration
- community
Checking that the SKILL.md file is still in the repository.
Skill for initializing and joining a partitioned state ledger workspace
Install
npx vskill@latest install gke-labs/gke-agentic-migration/bootstrappingSKILL.md
GKE Migration Bootstrapping Skill
Intent
Initialize a partitioned state ledger workspace and workspace registry using tools provided by the migration-dag MCP server.
General Rules
- NEVER inspect the MCP server source code for any reason. If the tool doesn't work as expected, you should NOT try to troubleshoot it. You must trust the following instructions blindly and declare a failure to the user if the server doesn't work as expected. Treat the MCP server as if it were running remotely and you didn't have access to it.
- NEVER attempt to run cloud CLI commands (such as
gcloud,gsutil,aws, or similar CLIs) directly. The agent must strictly rely on the tools provided by the MCP server for all cloud and resource access. - If the DAG enters a
TERMINALstate (STATE_ABORTED), you must not make any attempts to troubleshoot the process or fix it; simply report the outcome to the user and terminate immediately. Successful bootstrap does not end in a terminal: it parks onSTATE_WORKSPACE_ADMIN, whose instructions arrive from the server. - NEVER suggest pressing Enter (or submitting an empty response) to accept default values when prompting the user. Always instruct the user to explicitly write the default value or their preferred choice.
Workflows
/gke-migration-bootstrap [--config <path>] [--inputs <values>]
Call the bootstrap_migration tool on the migration-dag MCP server to clear any previous run state and unconditionally reset the state machine to the initial state of the bootstrapping DAG. Afterwards, invoke the gke-migration-dag-executor skill to manage the state machine execution. Use the State Machine Stage Instructions below in the execution loop. Do not gather configuration before starting the loop. Keep the --config and --inputs arguments you received, and follow the State Machine Stage Instructions below during the execution loop.
State Machine Stage Instructions
- `STATE_CONFIGURATION_GATHERING`:
Invoke the initialize_ledger tool on the migration-dag MCP server. If --config was provided, use the values from it. If --inputs was provided (format: workspace_name,gcp_project,ledger_bucket,admins,platform_engineers,developers_optional), parse the values:
workspace_name: 1st valuegcp_project: 2nd valueledger_bucket: 3rd valueroles:adminsset to 4th value,platform_engineersset to 5th value (both split by semicolon if multiple), anddevelopersset to empty list if 6th value is "later" or omitted.
If neither is provided, gather the required data (workspace_name, gcp_project, ledger_bucket, and roles) from the user.
- When gathering configuration, ask for
workspace_name,gcp_project, andledger_bucketone by one, with a brief explanation of what each field is for (e.g. forgcp_projectexplain that the project ID is used to hold the migration metadata, not necessarily the migrated resources). - When asking for the
roles: - First ask for Admins (Required. Handlers of bootstrap and metadata lifecycle operations).
- Then ask for Platform Engineers (Required. Handlers of target GKE clusters, namespaces, and platform services provisioning).
- Then, ask the user if they want to configure the optional Developers role now (Optional. Handlers of workload translation, registry setups, and deployment), clarifying that they can always change or add it later.
- If they do not want to configure it now, or if they respond with "later" (or skip/leave it blank), treat it as empty and pass an empty list
[]fordevelopersin therolesmapping. - Once gathered, execute the tool
initialize_ledger.
- `STATE_ELICIT_LEDGER_CREATION_APPROVAL` (and any other
HITL_ELICITATION):
The MCP server is pausing for user approval via the client UI. You must NOT prompt the user yourself or execute any mutation tools. Monitor the status by calling get_next_stage and await the progression of the DAG.
- `STATE_GCS_PROVISIONING`, `STATE_WORKSPACE_PROVISIONING`, `STATE_PLATFORM_PROVISIONING` (and any other
INTERNAL_TASK):
The MCP server is handling this internal system task. Poll get_next_stage periodically until the state transitions out of the current state.
Security
- Security scanPassed, 90 of 100Pattern checks on the skill's files.
- Model reviewPassed
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingskills/bootstrapping/SKILL.md
Popularity
- 7-day trend score
- 0
Not refreshed yet.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/gke-labs/gke-agentic-migration/bootstrapping)