Document the local checks and index the example packs (closes #31, closes #34)

CONTRIBUTING.md

@@ -41,6 +41,29 @@ A pull request should:
 
 Normative changes are not accepted solely because the reference implementation supports them.
 
+## Local checks
+
+CI runs the repository checks on Python 3.12. That is what the workflow currently uses rather
+than a declared minimum version for the project.
+
+Install the test dependencies:
+
+```bash
+python -m pip install -r requirements-dev.txt
+```
+
+Run the same checks CI runs:
+
+```bash
+python -m unittest discover -s tests -v
+```
+
+**Install the dependencies even if the suite appears to run without them.** `requirements-dev.txt`
+pins `jsonschema[format-nongpl]`, and the `format` keyword is only checked when those extras are
+present. Without them, `jsonschema` silently ignores `format`, so conformance cases that exist to
+prove a malformed URI or timestamp is rejected come back **valid** and the suite passes while
+checking less than it says it does. Two current corpus cases behave exactly that way.
+
 ## Developer Certificate of Origin
 
 Contributions must be signed off using Git's `--signoff` option. The sign-off certifies the

examples/README.md

@@ -0,0 +1,36 @@
+# Example packs
+
+Five synthetic Judgment Packs, checked in so a reader can see the format working before
+writing one. Each is a complete document that validates against
+[`schema/judgment-pack-core.schema.json`](../schema/judgment-pack-core.schema.json).
+
+| Example | Synthetic domain | Worth inspecting |
+| --- | --- | --- |
+| [`data-request-intake-triage.json`](data-request-intake-triage.json) | triaging an incoming request for data | the only pack with an **optional** evidence requirement (`required: false` on `sensitive-data-approvals`) and the only one using `attestation` evidence — a good place to see how an absent-but-optional requirement differs from a missing one |
+| [`minimal-expense-approval.json`](minimal-expense-approval.json) | approving an expense claim | the smallest complete pack here, and the clearest look at ordered comparison: `in`, `greater-than` and `less-than-or-equal` on one decision, with a `not` wrapping a nested condition |
+| [`records-disposition-review.json`](records-disposition-review.json) | deciding whether records may be disposed of | evidence-light by design — one requirement, no exceptions — so the rule structure is what carries the decision |
+| [`software-change-review.json`](software-change-review.json) | reviewing a proposed software change | two independent `evidence-present` conditions inside one `all`, which is the shape most real packs reach for first |
+| [`supplier-invoice-approval.json`](supplier-invoice-approval.json) | approving a supplier invoice | an `exception` alongside a `not` condition — the pair worth reading together, because an exception is not the same thing as a negated rule |
+
+## What these are, and are not
+
+They are **synthetic and non-operational**. Every organisation, threshold, role and source in
+them is invented for illustration. None of them is authorization, production policy, an
+industry standard, or evidence that any decision is safe or correct. A pack states what a
+decision was meant to consider; it does not make the decision right.
+
+Two things not to do with them:
+
+- **Do not add real or sensitive data.** These files are public, and a worked example is a
+  bad place for a real threshold, a real supplier, or anything about a real person.
+- **Do not fetch the source locators.** The `sources` entries carry example URIs and paths.
+  Ordinary validation reads the document and nothing else — no locator is dereferenced, and
+  a tool that fetched them would be doing something the specification does not ask for.
+
+## Where to go next
+
+- [`TESTING.md`](../TESTING.md) — how to validate a pack against the schema and the
+  conformance corpus.
+- [`docs/field-guide.md`](../docs/field-guide.md) — what each field means, in prose.
+- [`schema/judgment-pack-core.schema.json`](../schema/judgment-pack-core.schema.json) — the
+  normative shape these examples conform to.