Document the handoff to nomologR on the README, and fix the RC-day order

DESCRIPTION

@@ -1,7 +1,7 @@
 Package: contentvalidR
 Type: Package
 Title: Tools for Substantive and Content Validity Pretesting
-Version: 0.10.0
+Version: 0.10.0.9000
 Authors@R: c(
     person(given = "Joshua", family = "Uhalt", email = "Josh.Uhalt@gmail.com",
            role = c("aut", "cre"))

NEWS.md

@@ -1,3 +1,11 @@
+# contentvalidR 0.10.0.9000 (development version)
+
+* The README has a section on the handoff to nomologR, just above the
+  references: what a handoff carries, the two nomologR calls that read it, how
+  to read its decisions, and how the schema is versioned. The nomologR README
+  carries the mirror section, so the handoff can be found from either
+  package's front page.
+
 # contentvalidR 0.10.0
 
 Tenth public release, and the last planned before the joint 1.0 release

README.Rmd

@@ -144,6 +144,32 @@ Five deterministic example data sets, covering the item-sort, construct-rating,
 
 Cite the package with `citation("contentvalidR")`. The reference list below is also installed in BibTeX form: `system.file("REFERENCES.bib", package = "contentvalidR")`.
 
+## The handoff to nomologR
+
+Content review decides which items go on to be tested with response data. `content_handoff()` packages that decision so the empirical stage can pick it up without retyping anything, and [nomologR](https://github.com/JUhalt/nomologR) is the partner package that reads it. A handoff carries:
+
+- the items carried forward, and the items held back, each with its status, its recommendation, and the rule that made the decision;
+- the scales, where the design maps items to constructs;
+- each item's statistics, with their intervals and criteria, and the panel's agreement where one was computed;
+- two facts about the instrument that only you know: which items are reverse-worded (`reverse_keyed`, or `character(0)` if you checked and none is) and the response scale respondents will answer on (`response_scale`).
+
+```{r handoff}
+h <- content_handoff(fit, reverse_keyed = character(0), response_scale = c(1, 5))
+h$items
+```
+
+Once responses are collected, pass the handoff itself to nomologR, not `h$items`, so the keying and the reasons for anything held back travel with the items:
+
+```{r handoff-nomologr, eval=FALSE}
+library(nomologR)
+nomo_screen(responses, items = h)  # screens the carried items
+nomo_run(responses, scales = h)    # runs the empirical stage on the handoff's scales
+```
+
+**Reading a handoff.** Take each item's decision from `carried` and `status`; never re-derive it by comparing a statistic with its criterion, or by branching on the version that produced it. `recommendation` states the same decision in the workflow's own words and, like `rule`, is prose. A `keying` of `NA` means nobody said, never that an item is forward-worded.
+
+**Versions.** The handoff is schema version 1, frozen since 0.7.0: its fields and columns keep their names, positions, and types, and new optional ones may only be added at the end. A change that broke this would be schema version 2, produced beside version 1 for at least a release cycle. The reader in nomologR is tested against handoffs from contentvalidR 0.6.0 through 0.9.0, and neither package depends on the other. See `?content_handoff` for the full contract, `vignette("handoff-to-empirical-validation")` for a guide, and `vignette("one-item-set-both-stages")` for the joint walkthrough.
+
 ## References
 
 Works cited in this README, the help pages, and the vignettes.

README.md

@@ -267,6 +267,57 @@ Cite the package with `citation("contentvalidR")`. The reference list
 below is also installed in BibTeX form:
 `system.file("REFERENCES.bib", package = "contentvalidR")`.
 
+## The handoff to nomologR
+
+Content review decides which items go on to be tested with response
+data. `content_handoff()` packages that decision so the empirical stage
+can pick it up without retyping anything, and
+[nomologR](https://github.com/JUhalt/nomologR) is the partner package
+that reads it. A handoff carries:
+
+- the items carried forward, and the items held back, each with its
+  status, its recommendation, and the rule that made the decision;
+- the scales, where the design maps items to constructs;
+- each item’s statistics, with their intervals and criteria, and the
+  panel’s agreement where one was computed;
+- two facts about the instrument that only you know: which items are
+  reverse-worded (`reverse_keyed`, or `character(0)` if you checked and
+  none is) and the response scale respondents will answer on
+  (`response_scale`).
+
+``` r
+h <- content_handoff(fit, reverse_keyed = character(0), response_scale = c(1, 5))
+h$items
+#> [1] "Clear 1" "Clear 2"
+```
+
+Once responses are collected, pass the handoff itself to nomologR, not
+`h$items`, so the keying and the reasons for anything held back travel
+with the items:
+
+``` r
+library(nomologR)
+nomo_screen(responses, items = h)  # screens the carried items
+nomo_run(responses, scales = h)    # runs the empirical stage on the handoff's scales
+```
+
+**Reading a handoff.** Take each item’s decision from `carried` and
+`status`; never re-derive it by comparing a statistic with its
+criterion, or by branching on the version that produced it.
+`recommendation` states the same decision in the workflow’s own words
+and, like `rule`, is prose. A `keying` of `NA` means nobody said, never
+that an item is forward-worded.
+
+**Versions.** The handoff is schema version 1, frozen since 0.7.0: its
+fields and columns keep their names, positions, and types, and new
+optional ones may only be added at the end. A change that broke this
+would be schema version 2, produced beside version 1 for at least a
+release cycle. The reader in nomologR is tested against handoffs from
+contentvalidR 0.6.0 through 0.9.0, and neither package depends on the
+other. See `?content_handoff` for the full contract,
+`vignette("handoff-to-empirical-validation")` for a guide, and
+`vignette("one-item-set-both-stages")` for the joint walkthrough.
+
 ## References
 
 Works cited in this README, the help pages, and the vignettes.

ROADMAP.md

@@ -860,10 +860,17 @@ reads no handoffs and takes no part in the joint 1.0.
     another round is needed). It is a tag only, never a GitHub release,
     because R-universe publishes GitHub releases to users. Handoff fixtures
     from a candidate are named `handoff-<fit>-v0.99.0.rds`.
-  - **2026-10-17, release-candidate day.** `agreement_summary()` is removed,
-    the version becomes 0.99.0, and `v1.0.0-rc.1` is tagged. Then the round
-    below runs: the fixture set goes to nomologR, and the compatibility
-    table, the shared handoff example, and the release notes are drafted.
+  - **2026-10-17, release-candidate day,** for both packages, in this order,
+    agreed with nomologR so that its tests pass before either package tags:
+    1. contentvalidR removes `agreement_summary()`, sets the version to
+       0.99.0, pushes that commit untagged, and generates the fixture set
+       from it, with the commit recorded in the manifest;
+    2. nomologR adds the fixtures, reads 0.99.0 as a producer version, and
+       runs its suite;
+    3. once that passes, both packages tag `v1.0.0-rc.1`, contentvalidR on
+       exactly the commit the fixtures came from;
+    4. then the compatibility table, the shared handoff example, and the
+       release notes are drafted.
   - **CRAN submission, after that.** CRAN asks for updates to an established
     package no more often than every one to two months. So the joint 1.0 goes
     to CRAN no sooner than about a month after the later of the two