Build the **Interaction Landscape** research sprint

Added the one-shot Interaction Landscape research sprint — a survey of novel LLM interaction paradigms that renders as a draft briefing for review.

Below is the machine's own write-up. It's written for engineers and gets technical — the summary above is the whole story if you'd rather skip it.

Registered interaction-landscape in the EXPERIMENTS table in engine/run-experiment.mjs as a digest-kind experiment rendering type: briefing, and wrote engine/experiments/interaction-landscape.md — an operating manual mirroring Agent Weekly’s discipline (discovery-first broad search, worldwide sources, verify every fetched link, factual machine-log voice, no padding) but scoped to interaction paradigms across nine areas rather than releases, ending in a mandatory ranked ## Prototype Shortlist with each candidate marked client-only / BYO-key / needs-server. The title template hardcoded — week of <date>, which reads wrong for a one-shot, so it was generalized to an optional per-registry cadence word: cadenced experiments keep Agent Weekly — week of 2026-07-20 while the sprint titles as Interaction Landscape — 2026-07-20; the monitor dry-run preview now takes the same computed title instead of its own hardcoded copy. draftForType('briefing') already returned true via the fail-safe default, so no logic change was needed — the existing unit test was kept and annotated to record that Wolf’s review of the shortlist is the publishing gate. Verified node engine/run-experiment.mjs interaction-landscape --dry-run previews a type: briefing, draft: true entry with no side effects and that Agent Weekly and Site Health titles are unchanged; npm test (163 tests, 8 files) and npm run check (0 errors) both pass.