Privacy and data flows

OIS is a data format, and its tools run where you run them. ois-validator, ois-junit and ois-intent open no network connection: OIS records stay in your organization until you send them somewhere.

Who receives what

Step	What is transmitted	To	How to limit it
Installing ois-validator with uvx or pip	The download request: IP address and, as pip and uv document for PyPI’s download statistics, installer, Python and operating system version and whether it runs in CI	PyPI, run by the Python Software Foundation	Install from an internal package mirror, or a pinned wheel from your own artifact store
Installing uv with the setup-uv GitHub Action	The download request	GitHub	A runner image with uv preinstalled
Running ois-junit	Nothing. It reads the JUnit XML and the local Git checkout (git rev-parse, ls-files, show, remote get-url; never fetch or push) and writes a file	Nobody
Running ois-validator	Nothing. The JSON Schema ships inside the package and is never fetched	Nobody
Running ois-intent	Nothing. It reads and writes intent and decision files locally	Nobody
Storing bundles	The bundle file	Wherever your pipeline keeps artifacts	Your retention rules apply
Sharing a bundle or passport	Its records (below)	The recipients you choose, such as an auditor or a customer	Review before sharing (specification section 12)
Visiting this website	The request: IP address, page, browser	GitHub Pages, which hosts the site	The site loads nothing from other domains and sets no cookies; your theme choice stays in your browser
Opening an issue or pull request	What you write	GitHub, publicly	Report security issues privately
Writing to editors@openintentschema.io	Your e-mail	Proton Mail, the project’s mail provider

What a bundle from ois-junit contains

Contained	Not contained
Test names, as criterion statements	Test output, failure messages and stack traces
Paths of test files and the sha256 of their content	The content of test files or any source code
Repository URL without credentials, commit hash	User names, e-mail addresses, CI variables or secrets
Suite name, the namespace you chose, the pipeline run URL	Host names of build machines
Result per test and the time it was observed	Timing or performance data

Test names and repository URLs can still reveal internal information. Check them before a bundle leaves your organization; test names must not contain personal data (specification section 12.1).

How this was checked

ois-validator, ois-junit and ois-intent, and the validator’s whole test suite, were run in a sandbox that blocks all network access. Both work unchanged; a network request in that sandbox fails. What pip and uv send to PyPI is taken from their documentation, not measured here.

In the specification

Section 12 of the specification sets the privacy rules for records: no personal data apart from actor names and identifiers, pseudonyms for human actors, test data referenced by location and digest rather than embedded, and datasets with personal data classified as production or anonymized.