@sanning/proof 0.19.0 → 0.20.0

package/vectors/case-file-index.json

@@ -1,7 +1,7 @@
 {
   "rule": "case-file-index",
   "ruleVersion": 2,
-  "statement": "How a reader compares a case file's index, contents.json, with the folder it sits at the root of (evidence-bundle.md §5.6.3 v1.20, A25). The index is never an input to the case reading, and no finding here moves a verdict, a note, a contribution or an exit code. Findings come in this order: case_index_unreadable (and nothing else), case_index_pack_missing, case_index_pack_unlisted, case_index_file_missing, case_index_file_unlisted, case_index_file_differs, case_index_file_too_large, case_index_names_absent; within a kind by the UTF-8 bytes of the ref or path, and names_absent in the index's own order. Only files under a listed pack's folder or under report/ are reported as unlisted.",
+  "statement": "How a reader compares a case file's index, contents.json, with the folder it sits at the root of (evidence-bundle.md §5.6.3). The index is never an input to the case reading, and no finding here moves a verdict, a note, a contribution or an exit code. Findings come in this order: case_index_unreadable (and nothing else), case_index_pack_missing, case_index_pack_unlisted, case_index_file_missing, case_index_file_unlisted, case_index_file_differs, case_index_file_too_large, case_index_names_absent; within a kind by the UTF-8 bytes of the ref or path, and names_absent in the index's own order. Only files under a listed pack's folder or under report/ are reported as unlisted.",
   "howToUse": "Call readCaseFileIndex / read_case_file_index with the bytes of index_hex (null: no index in the folder) and `files`, and compare the result with `expected`. A finding lists only the fields it carries.",
   "rows": [
     {

package/vectors/case-split.json

@@ -1,5 +1,5 @@
 {
-  "description": "How `case` reads folders into packs: the verify page's splitting rule (sanning-io/verify-page src/lib/case.ts splitCase, src/lib/pack.ts assemble) over a walk whose rules are the command's own, and the bundle.json reader that verify and case share. Restated in both kernels and pinned here. Written by tools/gen-vectors/gen_case_split_vectors.py, whose expectations are literals, not kernel output. A tree row lists what to create under an empty folder <root>: a file (path, hex), a folder (dir), a symbolic link (symlink, its target as written), a file whose path is bytes (path_hex), or `files` empty files in a folder (fill). `args` are paths under <root>, and <root> in an expected error stands for that folder. File contents are lowercase hex.",
+  "description": "How `case` reads folders into packs: the verify page's splitting rule (splitCase, assemble) over a walk whose rules are the command's own, and the bundle.json reader that verify and case share. Restated in both kernels and pinned here. Written by tools/gen-vectors/gen_case_split_vectors.py, whose expectations are literals, not kernel output. A tree row lists what to create under an empty folder <root>: a file (path, hex), a folder (dir), a symbolic link (symlink, its target as written), a file whose path is bytes (path_hex), or `files` empty files in a folder (fill). `args` are paths under <root>, and <root> in an expected error stands for that folder. File contents are lowercase hex.",
   "limits": {
     "max_json_depth": 512,
     "max_walk_depth": 32,
@@ -230,7 +230,7 @@
       "args": [
         "pack.zip"
       ],
-      "description": "A .zip named on the command line is read as a case file (evidence-bundle §5.6, v1.20 A25): <name>.zip is read at <name>/<entry>, the folder unzip <name>.zip -d <name> makes, so the archive and its unpacked folder give one reading.",
+      "description": "A .zip named on the command line is read as a case file (evidence-bundle §5.6): <name>.zip is read at <name>/<entry>, the folder unzip <name>.zip -d <name> makes, so the archive and its unpacked folder give one reading.",
       "expected": {
         "not_regular": [],
         "outside": [],

package/vectors/chain-reading.json

@@ -11,7 +11,7 @@
     "null": "In prev_event_id: this record claims to begin its adapter chain.",
     "publicKey": "An alias, not a real Ed25519 key. The reader treats it as an opaque grouping token — a chain is one head per (public_key, chain_key), profile section 3.2 — so only its distinctness is load-bearing."
   },
-  "adapterNamespaces": "Verified against sanning-io/anchor @ main, 2026-09-07: `run` (LangChain callbacks, TS and Python — anchor#68 promoted this out of the old `langchain` namespace), `vercel_ai` (Vercel AI SDK middleware), `interchange` (Interchange adapter). All three carry `prev_event_id` and `seq`; their chain key is `root_run_id`, `chain_key` and `session_id` respectively. These are read ONLY where the record carries no chain of its own, and they are transitional: once every adapter sets the record's own previous_hash, a reader needs no adapter knowledge at all.",
+  "adapterNamespaces": "As written by the @sanning/anchor SDK: `run` (LangChain callbacks, TS and Python), `vercel_ai` (Vercel AI SDK middleware), `interchange` (Interchange adapter). All three carry `prev_event_id` and `seq`; their chain key is `root_run_id`, `chain_key` and `session_id` respectively. These are read ONLY where the record carries no chain of its own, and they are transitional: once every adapter sets the record's own previous_hash, a reader needs no adapter knowledge at all.",
   "recordScaffold": {
     "payload_version": 1,
     "spec_version": "sanning.events/v2",
@@ -165,7 +165,7 @@
     },
     {
       "id": "intact-adapter-chain-run",
-      "why": "the batched LangChain path writes its linkage only in metadata; its records all say GENESIS and carry no chain_key. Three of them must read as ONE chain with nothing to report — not as three roots of a fictional one. The namespace is `run`, not `langchain`: anchor#68 moved it, and a reader still pinned on the old name would find no chain here and report nothing, which is indistinguishable from a clean bundle.",
+      "why": "the batched LangChain path writes its linkage only in metadata; its records all say GENESIS and carry no chain_key. Three of them must read as ONE chain with nothing to report — not as three roots of a fictional one. The namespace is `run`, not `langchain`: a reader pinned on the name `langchain` would find no chain here and report nothing, which is indistinguishable from a clean bundle.",
       "events": [
         { "id": "e1", "publicKey": "kA", "previous_hash": "GENESIS", "context": {}, "metadata": { "run": { "run_id": "r1", "parent_run_id": null, "root_run_id": "root-1", "seq": 0, "prev_event_id": null } } },
         { "id": "e2", "publicKey": "kA", "previous_hash": "GENESIS", "context": {}, "metadata": { "run": { "run_id": "r2", "parent_run_id": "r1", "root_run_id": "root-1", "seq": 1, "prev_event_id": "@e1" } } },
@@ -180,7 +180,7 @@
     },
     {
       "id": "adapter-chain-one-event-dropped-run",
-      "why": "a deletion on the adapter path shows twice, and both are worth saying: the pointer dangles AND the sequence has a hole. anchor's own comment is that the slot is burned BEFORE the hash, so a dropped event consumed a seq number and leaves a permanent gap.",
+      "why": "a deletion on the adapter path shows twice, and both are worth saying: the pointer dangles AND the sequence has a hole. the adapter burns the slot BEFORE the hash, so a dropped event consumed a seq number and leaves a permanent gap.",
       "events": [
         { "id": "e1", "publicKey": "kA", "previous_hash": "GENESIS", "context": {}, "metadata": { "run": { "run_id": "r1", "parent_run_id": null, "root_run_id": "root-1", "seq": 0, "prev_event_id": null } } },
         { "id": "e2", "publicKey": "kA", "previous_hash": "GENESIS", "context": {}, "metadata": { "run": { "run_id": "r2", "parent_run_id": "r1", "root_run_id": "root-1", "seq": 1, "prev_event_id": "@e1" } } },