Trust and data flow¶

rup-flow has no account, no analytics and no service of its own. This page lists every boundary it touches.

On your machine the CLI reads and writes the project files and sends diagram sources to a PlantUML container on the loopback interface; the agent provider processes prompts and files under its own terms; a container registry is contacted only the first time the container starts

On your machine¶

It does	Where
Reads and writes files	Inside the folders you point it at: the documentation directory and `rup.config.json`
Sends diagram sources	To the PlantUML server you run locally, `http://127.0.0.1:8085` by default
Pulls a container image	The PlantUML image from its public registry, the first time you start the server

Nothing else is sent anywhere. rup-flow Pro adds one more image, the Mermaid renderer, pulled the first time you choose Mermaid.

The agent¶

The AI agent you use with the skill processes your prompts and the files it reads under its provider's terms and privacy policy, not rup-flow's. rup-flow cannot see or limit that. Choose an agent whose terms fit the material you document.

The PlantUML container¶

The bundled compose file publishes the container on 127.0.0.1 only and runs it with the SANDBOX security profile, which blocks file and address access from diagrams. Do not expose that port to a network without adding authentication. The CLI never uses an !include of a file or an address in the sources it writes.

Written output¶

Generated documentation contains whatever you or the code contain. Review it before you publish it, especially output derived from code in rup-flow Pro: paths, configuration and table names.

The explorer Pro¶

explorer.html carries a content security policy that blocks every external load, and inserts diagrams and documents as nodes, never as raw HTML. It can be shared as a file.

Reporting a vulnerability¶

Report vulnerabilities privately to the maintainer; do not open a public issue. The security policy explains where.