diff --git a/docs/xsts-deviations.md b/docs/xsts-deviations.md
new file mode 100644
index 0000000..44f1f59
--- /dev/null
+++ b/docs/xsts-deviations.md
@@ -0,0 +1,98 @@
+# XSTS deviation worklist
+
+This partitions the W3C XML Schema Test Suite (XSTS, 2006-11-06 archive)
+deviations by **root cause and kind**, so the burn-down is driven by
+understanding each rule rather than by chasing the count. First measured
+2026-06-13; baselined and ratcheted in `Tests/XSTSSuiteTests.swift`.
+
+## What the oracle measures
+
+The runner compares PureXML's verdict to the suite's expected validity. A
+schema is "accepted" iff `Schema.Document(source)` does not throw; an instance
+is "accepted" iff `validate` returns no errors. The four counts:
+
+| Kind | Meaning | Direction |
+|---|---|---|
+| valid schema rejected | legal XSD that PureXML refused to compile | too strict |
+| invalid schema accepted | illegal XSD that PureXML compiled anyway | too lenient |
+| valid instance rejected | valid document PureXML flagged | too strict |
+| invalid instance accepted | invalid document PureXML passed | too lenient |
+
+The oracle is binary and coarse: "invalid schema accepted" conflates *we have a
+bug* with *we do not implement this validity constraint by design*. The point of
+this document is to separate those.
+
+## Baseline (post #146 list-facet fix)
+
+| Kind | Count |
+|---|---|
+| valid schema rejected | 75 |
+| invalid schema accepted | 2467 |
+| valid instance rejected | 547 |
+| invalid instance accepted | 555 |
+
+XSTS is a deliberately adversarial, exhaustive corpus; mature validators
+(Xerces, libxml2) do not pass it fully. The goal is not a green number. It is to
+fix the deviations that are genuine correctness bugs affecting real documents,
+and to document the remainder as spec-justified, deliberate exclusions.
+
+## Category A: genuine correctness bugs (fix; real-data impact)
+
+These are cases where PureXML mishandles a feature it does support. Each is a
+real bug a user could hit with a well-formed schema and document.
+
+| Cluster(s) | Count | Root cause | Locus | Status |
+|---|---|---|---|---|
+| `list-NMTOKENS/IDREFS/ENTITIES` length facets | ~84 | length on a built-in list counted characters, not items | `XSDSimpleParser.simpleType` | **fixed (#146)** |
+| `union-*-pattern`, `union-*-enumeration` | ~200 | a union restricted by `enumeration`/`pattern` does not enforce the facet (value outside the enumeration is accepted) | `SimpleType.validateUnion` | open, one fix, high value |
+| `atomic-duration-minInclusive/maxInclusive` | ~50 | duration ordering bounds not enforced | duration comparison | open |
+| `reL`, `reI`, `RegexTest` (instance) | ~155 | missing Unicode block escapes (`\p{IsArmenian}` etc.) and category coverage in the regex engine | `Sources/Regex` | open, feature completion |
+| `atomic-QName-length`, `-minLength` | ~36 | character-length facet wrongly applied to `QName`; the spec/NIST treat QName length as non-constraining (Xerces agrees) | length facet on QName | open, over-strict |
+| `valueconstraint`, `isDefault` | ~26 | default/fixed value handling | element/attr defaults | open, needs sampling |
+| `particlesJk/Jj/Ha/Q/R` (schema), `particlesOb/Je` (instance) | ~110 | content-model construction/validation edge cases | content model | open, bounded |
+| `int_min/maxInclusive`, `int_min/maxExclusive` (schema) | ~16 | integer bound facets rejecting legal schemas | numeric facets | open, bounded |
+
+Recommended order: union facets (~200, one fix) → QName length (~36) → regex
+Unicode blocks (~155) → duration ordering (~50) → particles/int facets.
+
+## Category B: schema-for-schemas validity (unimplemented by design)
+
+The bulk of the 2467 "invalid schema accepted". PureXML compiles schemas
+permissively: it reads the components it understands and ignores structurally
+illegal content rather than validating the schema document against the
+schema-for-schemas. Examples found:
+
+- `ctB`: `Annotation information` (bare text content) is accepted.
+- `notatF`: `` placed inside `` (illegal position) is accepted.
+- `idA/idB/idC`: malformed identity-constraint definitions accepted.
+- `annotation/annotB`: illegal annotation placement accepted.
+- `addB`, `stF`, `ctG`, `ctH`: assorted illegal component structures accepted.
+
+| Cluster(s) | Count |
+|---|---|
+| `ctB/ctG/ctH` (complexType structure) | ~183 |
+| `idA/idB/idC` (identity constraints) | ~118 |
+| `notatF` (notation placement) | ~66 |
+| `annotation/annotB` | ~62 |
+| `addB`, `stF`, and the long tail | remainder of 2467 |
+
+**Decision:** enforcing this is a large, low-real-world-value effort (people do
+not author structurally broken schemas, and a too-lenient *compiler* does not
+corrupt valid-document validation). It would require a schema-document
+validation pass against the schema-for-schemas. Treated as a **documented,
+deliberate exclusion** unless a concrete need arises. If implemented, it is its
+own epic, not part of the instance-correctness burn-down.
+
+## Category C: arguable / needs investigation
+
+- `test`, `elemT` (invalid instance accepted, ~86): not yet sampled; classify before fixing.
+- QName length (listed in A) is spec-murky; it sits in A because NIST and Xerces agree the instance is valid, so rejecting it is the defensible bug.
+
+## How a fix lands
+
+Pick a Category A cluster, read 2-3 cases (schema + instance + expected) against
+the spec text, fix the owning code, add a focused regression test, re-run
+`swift test -c release --filter XSTS`, confirm the relevant count drops, and
+ratchet the baseline in `Tests/XSTSSuiteTests.swift` down to the new value.
+Tracking issues: #145 (invalid schema accepted), #146 (valid instance
+rejected), #147 (invalid instance accepted), #148 (valid schema rejected).