跳至正文
OpenAgentSkill
浏览已开放 · 目录记录日期 2026-11-21 · 自动更新和账户服务暂时关闭。
数据与分析

dbops-changeset

Generates Liquibase database changesets via a structured workflow: resolve schema+instance (from context or via API discovery), confirm DB engine type, generate Liquibase YAML, and present for HITL review with Accept / Deny / Accept & Commit actions. Use when asked to create, modify, or drop database tables, columns, indexes, or constraints as a Liquibase changeset, generate a database migration, or fix a failed migration. Supports update mode (refine existing changeset) and error-recovery mode (fix a failed migration).

Harness · 已有静态检查记录

安装与使用

运行命令前先阅读来源与依赖说明。复制命令不会自动安装或执行。

npx skills add harness/harness-skills --skill dbops-changeset

技能说明


name: dbops-changeset description: >- Generates Liquibase database changesets via a structured workflow: resolve schema+instance (from context or via API discovery), confirm DB engine type, generate Liquibase YAML, and present for HITL review with Accept / Deny / Accept & Commit actions. Use when asked to create, modify, or drop database tables, columns, indexes, or constraints as a Liquibase changeset, generate a database migration, or fix a failed migration. Supports update mode (refine existing changeset) and error-recovery mode (fix a failed migration). metadata: author: Harness version: 2.1.1 mcp-server: harness-mcp-v2 module: DBOPS summary: Generate, refine, review, and execute Liquibase changesets with Harness DBOPS license: Apache-2.0 compatibility: >- Requires Harness MCP v2 dbops toolset (resource types: database_schema, database_instance, database_default_authoring_instance, database_snapshot_object, database_execute_llm_authoring_pipeline). Connector lookup uses the connectors toolset connector resource. Review is done in chat (Accept / Deny / Accept & Commit).

DBOPS Changeset Generation

Generate Liquibase database changesets from natural language descriptions, anchored to a real schema and instance from the Harness DBOPS service.

When to Use

When the user asks to:

  • Create, add, modify, or drop tables, columns, indexes, constraints, or views as Liquibase changeset operations
  • Generate a database migration or changeset
  • Fix a failed migration or update an existing changeset
  • Apply schema changes as Liquibase YAML

Prerequisites

The Skill tool loads this file (SKILL.md) automatically — do not Read it again.

Before generating Liquibase YAML, Read reference.md using the exact path listed in the Skill tool result. It contains all Liquibase YAML rules, change types, rollback patterns, batch migration guidelines, MongoDB JSON string requirements, and the clarifying questions policy. Apply those rules in Step 4 when generating the changeset.

Full Workflow

Step 0: Check for Pre-Seeded Context (ALWAYS do this first)

At conversation start, the initial message may contain a JSON block with pre-selected identifiers injected by the Harness frontend UI:

{
  "dbops_schema_id": "users-schema",
  "dbops_instance_id": "prod-instance"
}

Read this JSON block (if present) before doing any tool calls.

What's in contextWhat to do
dbops_schema_id + dbops_instance_idSkip Steps 1 AND 2 — but immediately fetch instance metadata via harness_get(resource_type="database_instance", resource_id="<dbops_instance_id>", dbschema_id="<dbops_schema_id>") to obtain connector (needed in Step 3) and check hasSnapshot (needed in Step 3-B).
Only dbops_schema_idSkip Step 1 — use this schema ID. Then go to Step 2-A
NeitherFull Steps 1 and 2

Step 1: Discover Database Schemas (skip if dbops_schema_id is in context)

Call harness_list for the available schemas:

harness_list(resource_type="database_schema")

The response is a plain JSON array of objects. Each object has:

  • identifier — used as the schema ID in subsequent calls
  • name — human-readable name
  • migrationType — "Liquibase" or "Flyway" (only Liquibase supported)
  • instanceCount — number of linked instances
  • primaryDbInstanceId — the designated instance for LLM authoring (may be empty). If non-empty, Step 2-A uses this directly as the instance_id without further discovery.

Disambiguation:

  • If the array has one schema, select it automatically.
  • If the array has multiple schemas, use AskUserQuestion (selection mode): Show: [<name> (<identifier>) — <migrationType>, <instanceCount> instances] for each.
  • If the array is empty: "No database schemas found in this project. Create one in Harness DBOPS first."
  • If migrationType is "Flyway": "This skill supports Liquibase schemas only."

Step 2-A: Resolve Instance from Schema (when only dbops_schema_id is known)

When you have a schema_id but no instance_id (from context OR after Step 1 selection):

  1. Call harness_get to fetch the full schema:
 harness_get(resource_type="database_schema", dbschema_id="<schema_id>")
  1. Check the primaryDbInstanceId field in the response:
  • If set: use it as the instance_id. Inform the user: "Using primary instance <primaryDbInstanceId> for schema <name>." Then fetch the instance to obtain the connector field needed in Step 3:
harness_get(resource_type="database_instance", resource_id="<primaryDbInstanceId>",
            dbschema_id="<schema_id>")

Read connector from the response.

  • If empty or missing: fall through to Step 2-B.

Step 2-B: Auto-select Best Instance (when primaryDbInstanceId is empty)

Use the database_default_authoring_instance resource to let the server pick the best instance (oldest with a metadata snapshot; falls back to oldest overall):

harness_get(resource_type="database_default_authoring_instance", dbschema_id="<schema_identifier>")

The response is { identifier: string, name: string, hasSnapshot: boolean, connector: string }.

Handling the response:

  • If the tool call returns results: Use identifier as the instance_id and connector as the connector identifier for Step 3 (no separate instance detail fetch needed). Inform the user: "Using instance <name> (<identifier>) for schema <schema_name>." If hasSnapshot is true, the instance has metadata available for snapshot lookups — proceed to Step 3-B for any changeset intent (create, modify, or drop). If false, skip Step 3-B entirely.
  • **Not fou
阅读完整来源 ↗