Where do protocol test run records live? #44

ADR 0016 (#43) separates a test run from a review. A review is a named person's judgement of a protocol; a run is a fact about one execution: agent, model, date, which tests passed. The checker writes a run record (run.yaml, format in PROTOCOL_STANDARD.md) and the person attaches it wherever it is useful, usually the PR or issue they are working on.

That is enough while there are few runs. It does not answer "which agents and models is this protocol known to work with", which is the question worth answering once runs accumulate. Options, roughly in order of weight:

• A labelled issue per protocol in the content repository, run records posted as comments.
• A testing-log repository in the federation, one file per run, indexed like PROTOCOLS.yaml.
• A runs/ directory beside tests/ in the protocol directory. Rejected in the ADR: a commit and a PR for every run by every person, in a content repository, and runs conflated with the protocol's own history.

The benchmarking study in #5 is what will generate volume and should drive the choice. Nothing to decide until it does.

Drafted with Claude Code.