aiup-core

aiup-core is the stack-independent foundation of the AI Unified Process. It turns a product vision into requirements, an entity model, use cases, and test journeys that stack-specific plugins can implement.

Use this plugin for every AI Unified Process project. It works with any programming language because its output boundary is a set of reviewable files under docs/.

Skills and workflow

Inception          Elaboration                             Construction                            Transition
─────────────     ───────────────────────────────────     ───────────────────────────────────     ─────────────────────
/requirements  →  /entity-model  →  /use-case-diagram  →  /use-case-spec  →  /test-case        →  /acceptance-protocol
                                                    └→  /business-process  ↗                      /user-guide
                                                                                                  /release-notes

The last six skills drive the layer plugins (aiup-flyway, aiup-jooq, aiup-jpa, aiup-spring-boot, aiup-vaadin, aiup-angular, aiup-playwright), which require aiup-core 3.3.0 or later (aiup-playwright: 3.2.0): install one per layer of your project, run /stack once, then /use-case-implement UC-001, /use-case-test UC-001, and /use-case-coverage UC-001; a journey goes through /test-case-implement TC-001 and /test-case-coverage TC-001. /use-case-coverage delegates to the read-only uc-coverage-layered sub-agent. With a combined stack plugin, use its own /implement and test skills instead.

Each skill reads the artifacts created by earlier steps. The linked SKILL.md files are the authoritative reference for detailed inputs, outputs, and behavior.

Installation

Tessl

Initialize the project once:

tessl init --agent agents

agents is the vendor-neutral layout; use claude-code, cursor, gemini, codex, copilot, or copilot-vscode for a specific host. Then install the plugin:

tessl install ai-unified-process/aiup-core

Depending on the configured agent, skills may be exposed as slash commands or invoked by intent, for example "create the requirements catalog".

Claude Code

/plugin marketplace add ai-unified-process/marketplace
/plugin install aiup-core

See the marketplace installation guides for other agents and manual adoption.

Prerequisites

For the forward workflow, create docs/vision.md in the target project. It should describe the product mission, target users, goals, scope, and constraints. The marketplace provides a vision template.

Existing projects can begin with /reverse-engineer instead of a vision document.

Inputs and generated artifacts

your-project/
└── docs/
    ├── vision.md                    # maintained by the team
    ├── requirements.md              # /requirements
    ├── glossary.md                  # /requirements (domain terms and synonyms to avoid)
    ├── entity_model.md              # /entity-model
    ├── use_cases.puml               # /use-case-diagram
    ├── architecture/
    │   └── stack.md                 # /stack (layer plugins only)
    ├── processes/
    │   └── BP-001-*.bpmn            # /business-process (optional), read by /test-case
    ├── use_cases/
    │   └── UC-001-*.md              # /use-case-spec
    └── test_cases/
        └── TC-001-*.md              # /test-case

/spec-review checks these files against each other. Its bundled script spec_lint.py needs only Python and can run as a blocking CI gate; accepted findings of an existing project go into docs/.spec-lint-baseline.json. /spec-review report writes docs/spec-report.html: requirement coverage, the traceability matrix, and the findings on one page. /spec-review impact FR-014 lists what a change to a requirement, business rule, or use case affects before anything is edited, with the trace behind each hit and the skills that carry the change through.

Review and commit these files with the source code. Layer and stack plugins consume the entity model, use case specifications, and test journeys without depending on which coding agent produced them.

Project structure

aiup-core does not impose a source-code layout. Its specification skills write only to the documentation paths above and inspect an existing source tree when /reverse-engineer runs; /stack writes the stack file under docs/architecture/. The construction orchestrators /use-case-implement, /use-case-test, and /test-case-implement produce code and tests, but only by running the layer skills that docs/architecture/stack.md names — the layout is theirs, not core's. /use-case-coverage and /test-case-coverage only read. The installed layer plugins, or the selected combined stack plugin, detect or document the implementation layouts they support.

MCP servers

None. The layer and stack plugins bring the documentation servers their technologies need.

Related documentation

• Getting started
• Workflow and artifacts
• Project setup
• Writing a layer skill
• Choose a stack plugin

License

Apache-2.0 · © 2025-2026 Simon Martinelli and the AI Unified Process contributors

See LICENSE and NOTICE. "AI Unified Process" identifies the original methodology; derived works must retain the NOTICE file and must not present themselves as the official AI Unified Process.