fix(ci): repair link-check workflow + bump deps to zod 4

.github/workflows/links.yml

@@ -28,7 +28,14 @@ jobs:
       - name: Collect external URLs from sources
         run: |
           mkdir -p lychee
-          grep -rhoE 'https?://[^"'\''<>) ]+' data/ src/content/docs/ \
+          # Allow `)` inside URLs (Wikipedia disambiguation titles), then strip
+          # a single trailing `)` only when unbalanced (handles markdown
+          # `[text](url)`). Drop resolver templates containing `{placeholder}`.
+          # Restrict to YAML/MD so lockfiles inside data/ are not scanned.
+          grep -rhoE --include='*.yaml' --include='*.yml' --include='*.md' --include='*.mdx' \
+              'https?://[^]["'\''<>[:space:]]+' data/ src/content/docs/ \
+            | awk '{ n=gsub(/\(/,"&"); m=gsub(/\)/,"&"); if (m>n) sub(/\)$/,""); print }' \
+            | grep -v '[{}]' \
             | sed -E 's/[.,;:]+$//' \
             | sort -u > lychee/urls.txt
           echo "Collected $(wc -l < lychee/urls.txt) unique URLs"

CONTRIBUTING.md

@@ -26,12 +26,11 @@ A contribution does not create a claim to acceptance, prioritization, publicatio
 
 ## Review tracks
 
-Changes are routed to one of two tracks:
+Changes are routed to one of three tracks:
 
 - **Technical review** — typos, formatting, broken links, minor metadata, `last_checked` updates, uncontested aliases, build / tooling fixes. Needs automated validation and one technical reviewer.
 - **Expert review** — new works, new citation systems, new corpora, contested mappings, changes to deterministic ID inputs, status changes (`active` / `deprecated` / `withdrawn` / `blocked`). Needs technical validation, a documented rationale with sources, and at least one expert reviewer.
-
-The Board reserves decisions on legal or policy-sensitive matters (takedowns, blocking, licence policy).
+- **Board reservation** — takedowns, blocking, licence policy, and other legal or policy-sensitive matters. Decided by the Association Board.
 
 ```mermaid
 flowchart TD

decisions/ADR-0001-conforms-to-replaces-target-kind.md

@@ -0,0 +1,57 @@
+# ADR-0001: Replace `target_kind` with `dcterms:conformsTo`
+
+- **Status:** Proposed
+- **Date:** 2026-06-09
+- **Deciders:** @maehr
+- **Tags:** spec
+
+## Context and problem statement
+
+`MappingAssertion.target.target_kind` was introduced as an OPTIONAL human-readable scheme hint (`"cts"`, `"wikidata"`, `"doi"`, …). The spec already says validators MUST NOT key behaviour off it and the IRI in `identifier` is authoritative. Issue [#6](https://github.com/textrefs/textrefs.org/issues/6) observes that a field with no normative weight, accompanied by Appendix B's enumerated list of "known" scheme labels, is upkeep without payoff: every new scheme means another table row to police.
+
+Linked Art's [digital integration model](https://linked.art/model/digital/) handles the same problem with `conforms_to` pointing at the relevant specification (e.g. the IIIF profile URI), letting the IRI carry the conformance claim without a curated label registry.
+
+## Decision drivers
+
+- The field is already non-normative; the IRI is authoritative.
+- Appendix B's label column would otherwise grow indefinitely.
+- Pre-v1.0.0 — breaking field renames are still acceptable.
+- Want to align with established Linked Data practice rather than invent a TextRefs-specific convention.
+
+## Considered options
+
+1. **Replace `target_kind` with `conforms_to`** — drop the field, introduce an optional `target.conforms_to` typed as `dcterms:conformsTo` in the JSON-LD context, accepting an IRI or array of IRIs.
+2. **Deprecate `target_kind`, keep accepting it** — add `conforms_to` alongside, mark `target_kind` deprecated. Smoother for downstream consumers; carries the dead field into v1.
+3. **Leave `target_kind` as-is** — no change. Fails to address the upkeep concern that prompted the issue.
+
+## Decision
+
+We choose **Option 1**. The replacement happens before v1.0.0 freezes the schema, so a clean break is preferable to carrying a deprecated field into the stable surface. `target.conforms_to` is OPTIONAL; the registry will not synthesise values where none are known. Multiple IRIs are allowed via array form so a target can claim conformance to several specifications.
+
+In the JSON-LD context, `conforms_to` maps to `dcterms:conformsTo` with `@type: @id` so the value is treated as an IRI reference.
+
+## Consequences
+
+### Positive
+
+- One less curated label registry to maintain.
+- IRIs are self-describing; conformance claims are dereferenceable.
+- Aligns with Linked Art's established pattern.
+
+### Negative / trade-offs
+
+- Breaking rename for any downstream consumer that read `target_kind`. Acceptable pre-v1.0.0.
+- Coordinated change across two repos (`textrefs/textrefs.org` schema/docs, `textrefs/registry` YAML data).
+
+### Follow-up actions
+
+- [x] Update spec §10 and Appendix B.
+- [x] Update the v1 JSON-LD context.
+- [x] Update Zod schema, compile pipeline, and in-tree fixture.
+- [x] Migrate `data/works/*.yaml` in `textrefs/registry` (carried by registry PR #1).
+- [x] Bump `data/` submodule pointer to the migrated registry commit.
+
+## Links
+
+- Related issues / PRs: textrefs/textrefs.org#6, textrefs/textrefs.org#5, textrefs/registry#1
+- External references: [Linked Art — Digital Integration](https://linked.art/model/digital/), [DCMI Terms — conformsTo](http://purl.org/dc/terms/conformsTo)

lychee.toml

@@ -33,6 +33,7 @@ exclude = [
 	"^https?://([a-z0-9-]+\\.)*linkedin\\.com",
 	"^https?://([a-z0-9-]+\\.)*twitter\\.com",
 	"^https?://([a-z0-9-]+\\.)*x\\.com",
+	"^https?://([a-z0-9-]+\\.)*opencollective\\.com",
 	"^https?://localhost(:\\d+)?(/|$)",
 ]