devopsflow

Bootstrap DevOpsFlow Codex hook runtime assets into an installed plugin mirror before other hooks run.

使用事件风暴、CQRS 和需求追踪进行通用 DDD 领域建模。Use this skill when Codex needs to clarify raw business requirements, split stakeholder needs into requirement items, evolve a persistent domain model, identify actors and multi-role collaboration, domain events, commands, policies, aggregates, domain services, read models, produce structured Markdown or optional Mermaid/PlantUML diagrams, and review designs that may be CRUD, database, package-structure, or DDD-terminology driven instead of problem-domain driven. Especially use for CRUD-looking admin requirements such as company, department, position, employee, account, role, or permission management that need behavior-first modeling instead of flat noun aggregates.

Design and review strict Google Cloud IAM-style authorization catalogs for RBAC and ABAC across REST, GraphQL, gRPC, protobuf, and mixed API surfaces. Use when Codex needs to name permissions or roles, define role bindings and CEL conditions, map API operations to authorization checks, migrate legacy colon-delimited permissions, or prevent multiple authorization naming schemes.

Pure TDD workflow for greenfield feature development, bug fixes, behavior-preserving refactors, characterization tests, and characterize-then-fix work. Use when Codex needs test-first discipline to build new behavior from scratch or change existing code without drifting behavior. Applies across languages and architectures. Do not use for technology-specific layering rules, documentation-only edits, trivial formatting-only changes, or purely exploratory prototypes where no executable behavior is expected.

Design, review, and refactor resource-oriented HTTP, REST, gRPC transcoding, protobuf, and OpenAPI APIs using Google API Improvement Proposals (AIP). Use when Codex needs to create or critique API endpoints, REST URI version prefixes, resource names, standard methods, custom methods, pagination, filtering, field masks, errors, long-running operations, protobuf service definitions, or api-linter setup against Google AIP guidance.

Mandatory engineering workflow router. Use at the start of software development, refactoring, bug fixing, domain modeling, glue coding, API or authorization design, code review, verification, branch finishing, or commit preparation to classify the task and select required skills such as df-resumable-workflow-guard, df-ddd-event-storming-design, df-glue-coding, df-iam-access-control-design, df-ddd-to-tdd-handoff, df-implementation-planning, df-executing-implementation-plan, df-tdd-skill, df-spring-web-boundaries, df-systematic-debugging, df-requesting-code-review, df-receiving-code-review, df-verification-before-completion, df-finishing-development-branch, or df-parallel-agent-orchestration.

Glue Coding workflow for AI implementation that should reuse local project material without blindly copying old structure. Use after domain ambiguity has been resolved or declared thin, and before implementation planning or coding, when a task resembles existing CRUD pages, list/form/detail screens, imports/exports, endpoints, adapters, command handlers, projections, tests, configuration, refactors, migrations, modernization, or other repeatable project patterns. The skill discovers local style packs, reference code, examples, docs, AGENTS.md rules, and nearby production implementations, classifies them as target patterns, legacy patterns, anti-patterns, or behavior evidence, preserves the right conventions, and limits new code to the intended delta.

Protect long-running or interruption-prone engineering work with explicit checkpoints, resume files, and handoff notes. Use when a task may span multiple turns or sessions, exceed about 30 minutes, touch multiple workflow stages, require repeated verification, approach context limits, resume after interruption/compaction, or when the user says continue, resume, checkpoint, long task, handoff, 中断, 继续, 续跑, 检查点, 长任务, or 交接.

Convert confirmed DDD event-storming outputs and requirement traceability into executable, language-agnostic TDD implementation slices. Use after df-ddd-event-storming-design has produced confirmed requirements, commands, events, aggregates, policies, invariants, read models, or relationships and Codex needs tests, implementation planning, or development sequencing.

Draw.io / diagrams.net diagram authoring and maintenance. Use when Codex needs to create, edit, review, repair, or explain .df-drawio.svg files, .df-drawio files, mxfile XML, mxGraphModel, mxCell nodes/edges, architecture diagrams, flowcharts, sequence-like diagrams, data-flow diagrams, deployment diagrams, or repository documentation diagrams that should open cleanly in diagrams.net.

Execute an implementation plan safely. Use when Codex has a written plan or is about to follow planned engineering steps: first review the plan for missing tests, oversized steps, wrong ordering, or stale assumptions, then execute one task at a time with verification evidence.

Finish a development branch for commit, push, PR, or handoff. Use after implementation and verification to inspect git status, separate user-owned changes from Codex changes, confirm tests and plan completion, prepare a Chinese Conventional Commit message, and draft PR or handoff notes.

Write a concrete, small-step implementation plan before coding. Use for multi-step features, bug fixes, refactors, DDD-to-TDD handoffs, risky behavior changes, or any engineering task that needs files, commands, expected RED/GREEN results, verification steps, and completion criteria before implementation.

Java-specific Jimmer ORM guardrails. Use when Codex changes Java Jimmer SQL entities, APT/Maven/Gradle annotation processing, Spring Boot integration, generated table-class queries, fetchers, DTO language files, graph saves, repositories, or persistence tests. Use this for Java source sets and Java-first mixed JVM projects; use df-jimmer-kotlin for Kotlin source sets.

Kotlin-specific Jimmer ORM guardrails. Use when Codex changes Kotlin Jimmer SQL entities, KSP/Gradle setup, Spring Boot integration, generated table-object queries, fetchers, DTO language files, graph saves, repositories, or persistence tests. Use this for Kotlin source sets and Kotlin-first mixed JVM projects; use df-jimmer-java for Java source sets.

General Jimmer ORM workflow and language router. Use when Codex changes Jimmer entities, compile-time generated sources, SQL DSL queries, fetchers, DTO language files, graph saves, repositories, Spring Boot integration, or persistence tests in JVM projects. Continue with df-jimmer-kotlin for Kotlin-first Jimmer projects and df-jimmer-java for Java-first Jimmer projects.

Plan safe parallel agent work for engineering tasks. Use when independent modules, plan review, implementation, verification, or code review can run in separate contexts with non-overlapping file ownership, or when a fresh reviewer context is valuable to avoid self-confirming complex work.

Address code review feedback safely. Use when Codex receives PR comments, review threads, requested changes, inline code review feedback, or maintainer suggestions and must classify comments, decide what to fix or discuss, apply focused changes, verify each fix, and avoid unrelated refactors.

Use when creating, updating, executing, or closing GitHub milestones, release plans, release checklists, tagged releases, goal/roadmap issue sets, deployment-proof workflows, failure reviews, or process feedback loops that must keep issue scope, milestone descriptions, Mermaid diagrams, draw.io diagrams, templates, and workflow skills aligned.

Move human-only scripts, one-off maintenance helpers, local environment launchers, and Gradle init scripts out of high-salience repository roots. Use when cleaning agent context noise, relocating reset or run scripts into docs/skills, or converting Groovy init.gradle snippets to Kotlin DSL init.gradle.kts templates.

Prepare implemented code for review with a bug-focused self-review. Use after the main implementation is complete and before handoff, PR, or final completion to inspect diffs for behavioral regressions, missing tests, risky contracts, data consistency, security, error handling, and unclear verification evidence.

Spring Web boundary guardrails for Java/Spring Boot refactors. Use when Codex changes controllers, REST endpoints, request/response mapping, validation, security, file upload/download, exports, or service/controller layering in Spring applications. This is an optional architecture and technology-stack companion to TDD skills; do not use as a generic TDD methodology skill.

Evidence-based debugging workflow for failing tests, production bugs, regressions, flaky behavior, or unclear root causes. Use when Codex must reproduce a problem, locate the earliest wrong point, add observations or minimal tests, identify root-cause evidence, fix one cause at a time, and add regression coverage before completion.

Mandatory completion gate for engineering work. Use before saying a development, refactor, bug fix, modeling, review, documentation, configuration, commit, or PR task is complete to verify user requirements, changed files, commands run with exit codes, skipped checks, manual validation, residual risks, unrelated changes, and Chinese Conventional Commit readiness.