Split the RFC 0006 accepted set into focused child RFCs (6.1.1)

.config/nextest.toml

@@ -99,6 +99,15 @@ success-output = "immediate"
 # budget and the test's own allowance meet.
 slow-timeout = { period = "60s", terminate-after = 10 }
 
+[[profile.default.overrides]]
+# The RFC 0006 coverage contract (obligation COV-4 in
+# docs/execplans/6-1-1-split-rfc-0006-set-into-focused-child-rfcs-and-task.md).
+# This test's whole point is the line it prints: how many capability groups are
+# still unwritten. A half-finished split passes every other coverage check, so
+# a stall is only visible if the count reaches the terminal on a green run.
+filter = 'test(/^coverage_map_status_is_reported($|::)/)'
+success-output = "immediate"
+
 [profile.ci]
 # The CI-only profile. It exists solely to carry the whole-run budget, and
 # inherits everything else from `default`, including that profile's

.github/workflows/release-dry-run.yml

@@ -16,9 +16,9 @@ concurrency:
 jobs:
   release:
     uses: ./.github/workflows/release.yml
-    with:
-      dry-run: true
     # This workflow runs on pull requests, so it forwards no secrets. The
     # release workflow reads only `secrets.GITHUB_TOKEN`, which a called
     # workflow receives without forwarding, so `secrets: inherit` would
     # only widen what a pull request can reach.
+    with:
+      dry-run: true

.gitignore

@@ -18,6 +18,7 @@ __pycache__/
 .coverage
 .uv-cache/
 .uv-tools/
+uv.lock
 .pytest_cache/
 .ruff_cache/
 .typos-oxendict-base.json

Makefile

@@ -230,6 +230,10 @@ test: test-nextest doctest ## Run every Rust test with warnings treated as error
 test-nextest: check-build-tools ## Run all non-doctest Rust tests through cargo-nextest
 	$(GATE_RUSTFLAGS) $(CARGO) nextest run --workspace --all-targets --all-features $(NEXTEST_BUILD_JOBS) $(NEXTEST_TEST_JOBS)
 
+.PHONY: test-rfc-stdlib-coverage
+test-rfc-stdlib-coverage: check-build-tools ## Check the RFC 0006 split and its parsers
+	$(GATE_RUSTFLAGS) $(CARGO) nextest run --test rfc_stdlib_coverage_tests --all-features $(NEXTEST_BUILD_JOBS) $(NEXTEST_TEST_JOBS)
+
 doctest: check-build-tools ## Run doctests, which cargo-nextest cannot execute
 	$(GATE_RUSTFLAGS) $(CARGO) test --workspace --doc --all-features $(BUILD_JOBS)
 

docs/contents.md

@@ -75,6 +75,9 @@ operator, user, and contributor references are easier to find.
   Survey of the ansible-core Jinja standard library, with an explicit
   disposition for every candidate helper and Netsuke-native contracts for the
   accepted set.
+- [rfcs/0013-structured-data-interchange-helpers.md](rfcs/0013-structured-data-interchange-helpers.md):
+  First focused child of RFC 0006: the JSON and YAML interchange helpers and
+  the cross-cutting contract discharged for them.
 - [rfcs/0007-netsukefile-testing-framework.md](rfcs/0007-netsukefile-testing-framework.md):
   Proposed Netsukefile testing framework: the `netsuke test` command, the YAML
   test dialect, and its mocking model.
@@ -232,6 +235,10 @@ operator, user, and contributor references are easier to find.
   Runtime annotation introspection declared unsupported for the workflow
   contract tests, with the loader gate's stale `TYPE_CHECKING` claim corrected
   and a revisit gate that reopens on a real consumer.
+- [ADR-040](adr-040-focused-child-rfcs-for-survey-rfcs.md): Splitting a survey
+  RFC into focused child RFCs, each owning one or more capability groups, with
+  the accepted set partitioned by a coverage map and guarded by a
+  derivation-based coverage test.
 - [ADR-041](adr-041-canonical-recipe-shell-quoting-surface.md): `shell_quote`
   and `shell_join` as the canonical recipe quoting surface, with two dialects
   and a host-dependent default.

docs/developers-guide.md

@@ -4567,6 +4567,58 @@ provisions `pytest`, `pyyaml`, `hypothesis`, and `cmd-mox==0.2.0` through
 `uv run --with`, so `uv` is the only prerequisite and no virtual environment
 needs creating by hand.
 
+### RFC 0006 standard-library coverage contract
+
+`tests/rfc_stdlib_coverage_tests.rs` runs seven repository checks:
+
+| Check                                         | Responsibility                                      |
+| --------------------------------------------- | --------------------------------------------------- |
+| `every_accepted_helper_has_exactly_one_owner` | Partition accepted helpers across child RFCs.       |
+| `no_forbidden_helper_is_registered`           | Keep deferred and rejected candidates out.          |
+| `totals_and_purity_aggregate_agree`           | Reconcile accepted totals and purity counts.        |
+| `coverage_map_status_is_reported`             | Report written and unwritten capability groups.     |
+| `inter_document_links_resolve`                | Resolve relative RFC links to files in the tree.    |
+| `every_capability_has_a_roadmap_task`         | Match owned helpers to tasks in the assigned step.  |
+| `every_child_discharges_every_clause`         | Check child RFC clause coverage and section 5 text. |
+
+The private modules in `tests/rfc_stdlib_coverage/` parse the source documents,
+derive helper sets, and validate the partition. `survey`, `section7`,
+`section8`, `totals`, and `assertions` read RFC 0006; `map`, `registries`, and
+`clauses` read its coverage map and each child RFC; `roadmap` reads
+`docs/roadmap.md`; `links` checks relative targets across the RFC corpus.
+`document` and `markdown` provide the shared structural and lexical parsing
+used by those readers. The contract therefore depends on the Markdown in RFC
+0006, child RFCs, and the roadmap staying in the supported shapes.
+
+Partial coverage can pass: unwritten child groups are allowed while the split
+is in progress, and the status check reports rather than rejects that state.
+The output count is the liveness signal; it must reach zero before the split is
+complete. `.config/nextest.toml` gives
+`test(/^coverage_map_status_is_reported($|::)/)` the
+`success-output = "immediate"` override so this count appears during the suite,
+as soon as this test finishes, including on a successful run.
+
+Run `make test-rfc-stdlib-coverage` for the focused Rust contract binary,
+`make test-workflow-contracts` for the workflow contracts, and `make test` for
+the full Rust suite. The focused target uses the repository's standard Rust
+gate flags and Nextest configuration. The parser tests live beside the private
+functions they exercise; filesystem cases use isolated temporary fixtures via
+the test-only `Repo` constructor, which is private to that module tree. Those
+fixtures are not a general-purpose Markdown parsing API. The link property's
+private depth helper belongs only to its reference model: it checks traversal
+before the model renders the normalized path. The Nextest contract's private
+policy helper checks only an override already selected by the exact filter; its
+caller owns loading and selection. Neither helper is a shared parser or a
+runtime interface.
+
+The parser intentionally handles a narrow subset: ATX headings, simple
+pipe-delimited table rows, and fenced blocks with up to three leading spaces.
+It does not parse full CommonMark or recognize fences nested in list items.
+Relative links are extracted from inline Markdown link syntax and checked for
+target-file existence; fragment anchors are not validated. See
+[ADR-040](adr-040-focused-child-rfcs-for-survey-rfcs.md) for the split and its
+coverage contract rationale.
+
 ### Configuration-precedence regression tests
 
 The config-precedence ladder and display-policy domain are covered by three