Skip to content

How rup-flow works

The person asks an agent; the agent reads the skill, runs the CLI and writes the project files; the CLI draws diagrams through a local PlantUML server The person asks an agent; the agent reads the skill, runs the CLI and writes the project files; the CLI draws diagrams through a local PlantUML server

The skill

SKILL.md holds the operating rules and tells the agent which reference to read for each request. The references (methodology, domain-model, use-cases, diagram-conventions, structure) are loaded on demand so the agent spends its attention on your project, not on boilerplate.

The skill is written once in the open Agent Skills format. Agents that do not read that format read a small instruction file instead; see Compatible agents.

The CLI

A dependency-free Node.js program does the mechanical work:

It does Command
Creates the structure and the configuration init, new use-case
Keeps the use-case index in sync with the cards index
Draws .puml sources into .svg render
Verifies structure, links and diagrams check
Writes the block that agents without skills read instructions
Tells what is installed version, editions

The operating rules

  1. Read the configuration first.
  2. Mark every statement as confirmed, assumed or open.
  3. Never invent permissions, rules, states or attributes.
  4. Write where the configuration says.
  5. Do not claim a diagram is rendered or checked unless it was.
  6. Write documents in your language; keep identifiers in English.
  7. Keep one source of truth per artifact.

See Provenance marks and the architecture.

Editions

The same CLI runs both editions. rup-flow Pro adds an extension that the CLI loads when it is present; without it, the CLI is the free edition and says so when you ask for something outside it. Read Editions and the extension point.