feat(learning-harness): closed family taxonomy with a verifier per family

Task 1.1. Ten families are the only key that routes work and scores a pair;
CWE ids and mechanism ids are attributes reached through a family. Each family
names what can confirm a claim in it: a sandbox canary for injection, file path
and deserialization, static corroboration for access control, nothing for the
rest. The unknown family carries the mechanisms that span families and the
weaknesses this project deliberately does not target, so a declined row is
visible instead of silently unresolved.

Every labeled corpus weakness resolves, no weakness id is claimed twice, and
every mechanism in the closed vocabulary belongs to exactly one family. Open
redirect joins server-side request forgery under one untrusted-destination
family because both are untrusted input choosing a request destination; that
places 34 real rows instead of discarding them. The relation lookup answers
same, parent-child, lateral or fabricated, and abstention earns no partial
credit: unknown against a real family is lateral.

Co-Authored-By: Claude Opus 5 (1M context) <[email protected]>
Claude-Session: https://claude.ai/code/session_01T7XD7x5KonMdmGRkp63EQG

src/openultrasast/ruleset/families.toml

@@ -0,0 +1,126 @@
+# Closed family taxonomy (learning-harness, Req 1). A family is the ONLY key that routes work and
+# scores a pair; CWE ids and mechanism ids are attributes looked up through a family and are never
+# keys themselves. `verifier` says what can confirm a claim in this family: a canary oracle in the
+# sandbox, static corroboration from another checker, or nothing (the claim stays a suspicion).
+# `unknown` carries the generic mechanisms and the deliberately out-of-scope weaknesses, so a row we
+# have considered and declined is visible instead of silently unresolved. Bump `version` on any change.
+version = "1"
+
+[[family]]
+id = "injection"
+description = "Untrusted input reaches an interpreter: SQL, OS command, code, expression or server-side template."
+cwes = ["CWE-77", "CWE-78", "CWE-88", "CWE-89", "CWE-90", "CWE-91", "CWE-94", "CWE-95", "CWE-564", "CWE-917", "CWE-943", "CWE-1336"]
+mechanisms = []
+verifier = "canary"
+
+[[family]]
+id = "path"
+description = "Untrusted input reaches a filesystem path: traversal, unsafe join, file inclusion."
+cwes = ["CWE-22", "CWE-23", "CWE-36", "CWE-73", "CWE-98", "CWE-99"]
+mechanisms = ["path_join_user_input"]
+verifier = "canary"
+
+[[family]]
+id = "deserialization"
+description = "Untrusted data is deserialized into objects the process then trusts."
+cwes = ["CWE-502"]
+mechanisms = []
+verifier = "canary"
+
+[[family]]
+id = "access_control"
+description = "An operation on a protected resource is reached without the guard or ownership constraint its siblings carry."
+cwes = ["CWE-284", "CWE-285", "CWE-306", "CWE-425", "CWE-566", "CWE-639", "CWE-862", "CWE-863", "CWE-1220"]
+mechanisms = [
+    "identity_from_request_body",
+    "missing_auth_guard",
+    "unconstrained_protected_read",
+    "unconstrained_protected_write",
+    "unguarded_privileged_action",
+]
+verifier = "static"
+
+[[family]]
+id = "output_encoding"
+description = "Untrusted data reaches a rendered document or a header without encoding: cross-site scripting, response splitting."
+cwes = ["CWE-79", "CWE-80", "CWE-83", "CWE-113", "CWE-116"]
+mechanisms = []
+verifier = "none"
+
+[[family]]
+id = "untrusted_destination"
+description = "Untrusted input controls where a request or a redirect goes: server-side request forgery and open redirect."
+cwes = ["CWE-441", "CWE-601", "CWE-611", "CWE-918"]
+mechanisms = []
+verifier = "none"
+
+[[family]]
+id = "prototype"
+description = "Untrusted keys reach an object merge or attribute assignment and change behaviour other code relies on."
+cwes = ["CWE-471", "CWE-915", "CWE-1321"]
+mechanisms = ["prototype_pollution"]
+verifier = "none"
+
+[[family]]
+id = "config_secrets"
+description = "A security setting ships permissive or a credential ships in the artifact: CORS, debug, cookie flags, weak crypto, hardcoded secrets."
+cwes = [
+    "CWE-16",
+    "CWE-215",
+    "CWE-259",
+    "CWE-311",
+    "CWE-312",
+    "CWE-315",
+    "CWE-326",
+    "CWE-327",
+    "CWE-346",
+    "CWE-489",
+    "CWE-522",
+    "CWE-614",
+    "CWE-798",
+    "CWE-942",
+    "CWE-1004",
+]
+mechanisms = ["config_sink_weak_literal", "permissive_default", "secret_in_client"]
+verifier = "none"
+
+[[family]]
+id = "memory"
+description = "Memory safety in C and C++: bounds, lifetime, arithmetic. Measured on the vendored slice, never gated; specialized systems own this class."
+cwes = [
+    "CWE-119",
+    "CWE-120",
+    "CWE-121",
+    "CWE-122",
+    "CWE-125",
+    "CWE-126",
+    "CWE-131",
+    "CWE-134",
+    "CWE-190",
+    "CWE-191",
+    "CWE-401",
+    "CWE-415",
+    "CWE-416",
+    "CWE-476",
+    "CWE-680",
+    "CWE-770",
+    "CWE-787",
+    "CWE-843",
+]
+mechanisms = [
+    "fixed_buffer_capacity",
+    "null_deref_unguarded",
+    "signed_size_operand",
+    "type_confusion",
+    "uaf",
+    "unchecked_alloc_size",
+    "unchecked_length_copy",
+]
+verifier = "none"
+
+[[family]]
+id = "unknown"
+description = "No family admitted: a mechanism that spans families, or a weakness this project deliberately does not target."
+cwes = ["CWE-20", "CWE-835"]
+mechanisms = ["container_taint", "cross_artifact", "other", "redos", "source_reaches_sink", "validation_strength"]
+verifier = "none"