@dreamforgestudiollc/emet 0.3.5 → 0.3.6

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
package/docs/SETUP.md CHANGED
@@ -251,9 +251,15 @@ it should come back `ready` with your existing identity, not `setup_required`. I
251
251
  `setup_required`, one of the four variables above is wrong; **do not run the setup interview,
252
252
  which would write a new identity over your existing one.**
253
253
 
254
+ ## Upgrading to 0.3.6
255
+
256
+ 0.3.6 is a docs and wording release. Stored data shapes are unchanged. No new setting. Deploy over 0.3.5 as is. The public roadmap's Upcoming work section is not built unless it says so.
257
+
258
+ ISO 9001 7.5.3 / ISO 13485 4.2.5 document-control wording is unchanged.
259
+
254
260
  ## Upgrading to 0.3.5
255
261
 
256
- 0.3.5 brings the public docs current. Stored data shapes are unchanged. The bound door's tag is `bound-host`. Writes from a stored legacy token or sub-key are stamped `bound-host`. The alias covers stored data only. The previous spelling is refused for a new consent, a new sub-key request, a new writer request, a new member or guest code, a new setup source, and a chooser decision.
262
+ 0.3.5 brings the public docs current. Stored data shapes are unchanged. The bound door's tag is `bound-host`. Writes from a stored legacy token or sub-key are stamped `bound-host`. The alias covers stored data only. The previous spelling is refused for a new consent, a new sub-key request, a new writer request, a new member or guest code, a new setup source, a chooser decision, and a break-glass reissue.
257
263
 
258
264
  ISO 9001 7.5.3 / ISO 13485 4.2.5 document-control wording is unchanged.
259
265
 
package/package.json CHANGED
@@ -1,7 +1,7 @@
1
1
  {
2
2
  "name": "@dreamforgestudiollc/emet",
3
- "version": "0.3.5",
4
- "description": "ISO 9001 7.5.3 / ISO 13485 4.2.5 document-control memory for AI agents. Nothing is deleted. EMET 0.3.5 — Enduring Memory & Epistemic Tiers. An MCP memory server with persistent shared memory for AI agents, bot teams and chats: layered memory, verified writes, a verbatim transcript tier, per-client attribution and approvals, and client-side GEM execution. EMET verifies the record, not the computation. A GEM is seated only when its released flag is true. Every error carries a stable code; the owner-only emet_error_report returns a report and an email draft the user sends. npm run verify drives the registered tools from a checkout, and docs/features is the map of every feature.",
3
+ "version": "0.3.6",
4
+ "description": "ISO 9001 7.5.3 / ISO 13485 4.2.5 document-control memory for AI agents. Nothing is deleted. EMET 0.3.6 — Enduring Memory & Epistemic Tiers. An MCP memory server with persistent shared memory for AI agents, bot teams and chats: layered memory, verified writes, a verbatim transcript tier, per-client attribution and approvals, and client-side GEM execution. EMET verifies the record, not the computation. A GEM is seated only when its released flag is true. Every error carries a stable code; the owner-only emet_error_report returns a report and an email draft the user sends. npm run verify drives the registered tools from a checkout, and docs/features is the map of every feature.",
5
5
  "keywords": [
6
6
  "agent-memory",
7
7
  "ai-agents",
@@ -48,6 +48,9 @@
48
48
  "docs/*.md",
49
49
  "CHARTER.md",
50
50
  "CHANGELOG.md",
51
+ "RELEASE_NOTES*.md",
52
+ "SECURITY.md",
53
+ "CONTRIBUTING.md",
51
54
  "server.json",
52
55
  ".env.example",
53
56
  "README.md",
package/server.json CHANGED
@@ -3,7 +3,7 @@
3
3
  "name": "io.github.dreamforgestudiollc/emet",
4
4
  "title": "EMET",
5
5
  "description": "ISO 9001 7.5.3 / ISO 13485 4.2.5 document-control memory for AI agents. Nothing is deleted.",
6
- "version": "0.3.5",
6
+ "version": "0.3.6",
7
7
  "websiteUrl": "https://dreamforgestudiollc.github.io/EMET/",
8
8
  "repository": {
9
9
  "url": "https://github.com/dreamforgestudiollc/EMET",
@@ -22,7 +22,7 @@
22
22
  "registryType": "npm",
23
23
  "registryBaseUrl": "https://registry.npmjs.org",
24
24
  "identifier": "@dreamforgestudiollc/emet",
25
- "version": "0.3.5",
25
+ "version": "0.3.6",
26
26
  "transport": {
27
27
  "type": "stdio"
28
28
  },
@@ -132,6 +132,7 @@ const ANNOTATION_TABLE = {
132
132
  };
133
133
 
134
134
  function subkeyAnnotationsHidden() {
135
+ // CASCADE_SUBKEYS is a legacy environment name. EMET_SUBKEYS is the name to set.
135
136
  return process.env.EMET_SUBKEYS !== '1' && process.env.CASCADE_SUBKEYS !== '1';
136
137
  }
137
138
 
@@ -32,7 +32,7 @@
32
32
  * (transcript parts, documents, layer entries): if those rows carry another
33
33
  * tag, the write is refused.
34
34
  *
35
- * OWNER (A2). Live owner tokens carry the owner's person id (e.g. "alex");
35
+ * OWNER (A2). Live owner tokens carry the owner's person id;
36
36
  * older ones carry "owner" or nothing. All three are the owner.
37
37
  */
38
38
 
@@ -6,7 +6,7 @@
6
6
  * match, the close stays blocked, and retries used to count toward the
7
7
  * three-close cap.
8
8
  *
9
- * A records/ session disposition (templates/SESSIONDISPOSITION.md, #95)
9
+ * A records/ session disposition (templates/SESSIONDISPOSITION.md)
10
10
  * documents that fact. The row does not close the graph and it is not a
11
11
  * receipt. This path closes the graph only when the owner door, or the
12
12
  * credential that wrote the session, cites that current row and the later
package/src/database.js CHANGED
@@ -454,7 +454,7 @@ export class EmetDatabase {
454
454
  }
455
455
 
456
456
  /**
457
- * THE layer insert, native (2026-09-27, code audit 1.7 stage 1 - Alex: "if SQL
457
+ * THE layer insert, native (2026-09-27, code audit 1.7 stage 1 - the owner: "if SQL
458
458
  * is the old and needs to be updated then update it"). Takes the document as
459
459
  * a document. Every key must be a known layer field; the id is allocated
460
460
  * here. The INSERT-string shim it replaced went with the SQL translator in
package/src/gem-check.js CHANGED
@@ -96,6 +96,9 @@ const SECRET_PATTERNS = Object.freeze([
96
96
  /(api[_-]?key|secret|token|password)\s*[:=]\s*['"][^\s'"]{12,}/i
97
97
  ]);
98
98
 
99
+ /** In-repo schema id. Not a URL. The bundled schema and this check use the same string. */
100
+ export const GEM_SCHEMA_ID = 'emet:gem-library/gem.schema.json';
101
+
99
102
  const SECTION_KEY = /^[a-z][a-z0-9_]{0,40}$/;
100
103
  const FORBIDDEN_SECTIONS = new Set(['voice', 'engagement', 'persistence', 'corpus_discipline', 'operating_discipline']);
101
104
 
@@ -114,7 +117,7 @@ function loadSchema(explicit) {
114
117
  if (schemaCache) return schemaCache;
115
118
  const text = fs.readFileSync(schemaPath(), 'utf8');
116
119
  const parsed = JSON.parse(text);
117
- if (!parsed || parsed.$id !== 'https://github.com/dreamforgestudiollc/EMET-GEMs/raw/main/gem.schema.json') {
120
+ if (!parsed || parsed.$id !== GEM_SCHEMA_ID) {
118
121
  throw new CodedError('schema id', { code: 'EMET-GEM-013' });
119
122
  }
120
123
  schemaCache = parsed;
package/src/grants.js CHANGED
@@ -1,3 +1,5 @@
1
+ import crypto from 'node:crypto';
2
+
1
3
  /**
2
4
  * Door grants (R2 design scaffolding, 2026-10-26).
3
5
  *
@@ -140,8 +142,29 @@ export const MEMBER_REVOKE = Object.freeze({
140
142
  path: '/owner/break-glass/revoke-member'
141
143
  });
142
144
 
143
- /** The owl MCP clone account. Exact string. A substring is not this account. */
144
- export const OWL_ACCOUNT = 'user-EMET--owl';
145
+ /**
146
+ * Public label for logs and the retirement marker. Not compared to stored
147
+ * client ids. The retired account id is matched by digest so this package
148
+ * does not spell it. A substring is not that account.
149
+ */
150
+ export const OWL_ACCOUNT_LABEL = 'retired-owl-account';
151
+ /** sha256 of the retired account id, utf8. Exact match only. */
152
+ export const OWL_ACCOUNT_DIGEST = 'd2f8d11e0e331afd1315642e613b3f2d4c864904767c0a420761a40f81e01b0f';
153
+
154
+ export function isRetiredOwlAccount(value) {
155
+ if (typeof value !== 'string' || value.length === 0) return false;
156
+ const digest = crypto.createHash('sha256').update(value, 'utf8').digest('hex');
157
+ if (digest.length !== OWL_ACCOUNT_DIGEST.length) return false;
158
+ return crypto.timingSafeEqual(Buffer.from(digest), Buffer.from(OWL_ACCOUNT_DIGEST));
159
+ }
160
+
161
+ export function owlRetirementFailureLine(active = 'unknown') {
162
+ return `[emet] OWL RETIREMENT CHECK FAILED account=${OWL_ACCOUNT_LABEL} active=${active} rung=R6 pointer=/owner/chooser`;
163
+ }
164
+
165
+ export function owlRetirementPassLine() {
166
+ return `[emet] owl-retirement check=pass account=${OWL_ACCOUNT_LABEL} active=0 rung=R6`;
167
+ }
145
168
 
146
169
  /**
147
170
  * R6. The owl MCP clone account was a second OAuth client used only to get a
@@ -153,7 +176,7 @@ export const OWL_KILL = Object.freeze({
153
176
  rung: 'R6',
154
177
  implemented: true,
155
178
  delete_wiring: false,
156
- account: OWL_ACCOUNT,
179
+ account: OWL_ACCOUNT_LABEL,
157
180
  status: 'retired',
158
181
  superseded_by: 'R6',
159
182
  retained: true,
@@ -170,9 +193,9 @@ export const OWL_KILL = Object.freeze({
170
193
  export const WRITER_VALIDATOR = Object.freeze({
171
194
  rung: 'R6a',
172
195
  implemented: 'server',
173
- alex_choice_open: true,
196
+ owner_choice_open: true,
174
197
  options: Object.freeze(['server', 'existing-reviewer-credential']),
175
- not_an_option: OWL_ACCOUNT,
198
+ not_an_option: OWL_ACCOUNT_LABEL,
176
199
  note: 'The server-side check is what runs. the owner can later name another existing reviewer credential. The owl account is not a choice.'
177
200
  });
178
201
 
package/src/host-reach.js CHANGED
@@ -5,7 +5,7 @@
5
5
  * Licensed under the MIT License - see LICENSE
6
6
  *
7
7
  * The RUNE load gate lives in verification/load-gate/load-gate.mjs so
8
- * EMET-GEMs can vendor that file and its corpus byte-for-byte. This module
8
+ * A library checkout can vendor that file and its corpus byte-for-byte. This module
9
9
  * is the import path gem-check already uses. There is no second scan.
10
10
  */
11
11
 
@@ -33,6 +33,7 @@ import { requireBearerAuth } from '@modelcontextprotocol/sdk/server/auth/middlew
33
33
  import { Server } from '@modelcontextprotocol/sdk/server/index.js';
34
34
  import { StreamableHTTPServerTransport } from '@modelcontextprotocol/sdk/server/streamableHttp.js';
35
35
  import { CallToolRequestSchema, ListToolsRequestSchema } from '@modelcontextprotocol/sdk/types.js';
36
+ import { owlRetirementFailureLine } from './grants.js';
36
37
  import { EMET_VERSION } from './version.js';
37
38
 
38
39
 
@@ -349,7 +350,7 @@ if (process.env.EMET_HTTP_LISTEN !== '0') (async () => {
349
350
  await runStartupChecks(dbManager.db, { prefix: provider.prefix, log: (line) => console.error(line) });
350
351
  }
351
352
  } catch {
352
- console.error('[emet] OWL RETIREMENT CHECK FAILED account=user-EMET--owl active=unknown rung=R6 pointer=/owner/chooser');
353
+ console.error(owlRetirementFailureLine());
353
354
  }
354
355
  try {
355
356
  const { proofStartupLine } = await import('./proof-run.js');
package/src/index.js CHANGED
@@ -25,7 +25,7 @@
25
25
 
26
26
  import { Server } from "@modelcontextprotocol/sdk/server/index.js";
27
27
  import { setting } from './env.js';
28
- import { toolCallAuditDetails } from './grants.js';
28
+ import { owlRetirementFailureLine, toolCallAuditDetails } from './grants.js';
29
29
  import { redactText } from './redact.js';
30
30
  import { StdioServerTransport } from "@modelcontextprotocol/sdk/server/stdio.js";
31
31
  import {
@@ -267,7 +267,7 @@ async function main() {
267
267
  const { runStartupChecks } = await import('./owl.js');
268
268
  await runStartupChecks(dbManager.db, { log: (line) => logger.error(line) });
269
269
  } catch {
270
- logger.error('[emet] OWL RETIREMENT CHECK FAILED account=user-EMET--owl active=unknown rung=R6 pointer=/owner/chooser');
270
+ logger.error(owlRetirementFailureLine());
271
271
  }
272
272
  }
273
273
  try {
package/src/invites.js CHANGED
@@ -87,8 +87,8 @@ export async function createMemberAccess(db, { created_by, display_name, days =
87
87
  * A redeemed invite always gets a NEW person (multi-user report v2, defect 2,
88
88
  * 2026-10-01). Until then the guest's id was the slug of the display name the
89
89
  * invite carried, and an existing person with that slug was silently reused:
90
- * an invite named "Alex" (or "ALEX", "Aléx") redeemed as the owner, wrote as
91
- * the owner and passed the owner-only invite check. A display name is a label,
90
+ * an invite named "Member" (or "MEMBER", "Mémber") redeemed as the owner, wrote as
91
+ * the owner, and passed checks that only the owner should pass. A display name is a label,
92
92
  * not a credential, so it never selects an existing identity:
93
93
  * - the slug is a starting point; if any person already holds it, or it is a
94
94
  * reserved subject ('owner', the legacy owner subject and slugPersonId's
package/src/owl.js CHANGED
@@ -16,7 +16,7 @@
16
16
 
17
17
  import { COLLECTION_PREFIX } from './database.js';
18
18
  import { assignCode, newCorrelationId } from './error-codes.js';
19
- import { OWL_ACCOUNT, doorFromIssuedVia, grantIdFromFamily, logClientId, logDoor, logGrantId } from './grants.js';
19
+ import { OWL_ACCOUNT_LABEL, doorFromIssuedVia, grantIdFromFamily, isRetiredOwlAccount, logClientId, logDoor, logGrantId, owlRetirementFailureLine, owlRetirementPassLine } from './grants.js';
20
20
  import { InvalidGrantError, InvalidTokenError } from '@modelcontextprotocol/sdk/server/auth/errors.js';
21
21
 
22
22
  const POINTER = 'this owl MCP clone account was retired by R6. A new writer is prepared by the hub login, validated on the server, and decided by the owner at /owner/chooser. This account cannot approve. Records are retained.';
@@ -39,7 +39,7 @@ export function _resetOwlForTests() {
39
39
  export function owlRetirementMetric() {
40
40
  return {
41
41
  rung: 'R6',
42
- account: OWL_ACCOUNT,
42
+ account: OWL_ACCOUNT_LABEL,
43
43
  status: 'retired',
44
44
  retained: true,
45
45
  deleted: false,
@@ -56,13 +56,13 @@ export function owlRetirementMetric() {
56
56
  /** Exact account identity. A substring is not a match. */
57
57
  export function isOwlIdentity(value) {
58
58
  if (value == null) return false;
59
- if (typeof value === 'string') return value === OWL_ACCOUNT;
59
+ if (typeof value === 'string') return isRetiredOwlAccount(value);
60
60
  if (typeof value !== 'object') return false;
61
61
  const fields = [
62
62
  value.clientId, value.client_id, value.sourceTag, value.source_tag,
63
63
  value.effectiveTag, value.client_name, value.clientName, value.tag
64
64
  ];
65
- return fields.some((field) => field === OWL_ACCOUNT);
65
+ return fields.some((field) => isRetiredOwlAccount(field));
66
66
  }
67
67
 
68
68
  function notRetired(doc) {
@@ -91,11 +91,12 @@ async function readAll(db, name) {
91
91
  }
92
92
 
93
93
  /**
94
- * The account id, plus any registered client whose name or id is that
95
- * account. A token may carry an opaque registration id.
94
+ * Client ids of registered OAuth clients that are the retired account.
95
+ * A token may carry only an opaque registration id. The account string
96
+ * itself is matched by isOwlIdentity and is not stored in this set.
96
97
  */
97
98
  export async function owlClientIds(db, prefix = COLLECTION_PREFIX) {
98
- const ids = new Set([OWL_ACCOUNT]);
99
+ const ids = new Set();
99
100
  const clients = await readAll(db, `${prefix}oauth_clients`);
100
101
  for (const client of clients) {
101
102
  if (client && isOwlIdentity(client) && typeof client.client_id === 'string' && client.client_id) {
@@ -140,7 +141,7 @@ export async function ensureOwlRetirementMarker(db, prefix = COLLECTION_PREFIX,
140
141
  if (existing) return existing;
141
142
  const row = {
142
143
  _id: MARKER_ID,
143
- account: OWL_ACCOUNT,
144
+ account: OWL_ACCOUNT_LABEL,
144
145
  status: 'retired',
145
146
  superseded_by: 'R6',
146
147
  rung: 'R6',
@@ -167,8 +168,8 @@ export async function owlStartupCheck(db, prefix = COLLECTION_PREFIX, now = new
167
168
  }
168
169
  const ok = !failed && active.length === 0;
169
170
  const line = ok
170
- ? `[emet] owl-retirement check=pass account=${OWL_ACCOUNT} active=0 rung=R6`
171
- : `[emet] OWL RETIREMENT CHECK FAILED account=${OWL_ACCOUNT} active=${failed ? 'unknown' : active.length} rung=R6 pointer=/owner/chooser`;
171
+ ? owlRetirementPassLine()
172
+ : owlRetirementFailureLine(failed ? 'unknown' : active.length);
172
173
  metric.ok = ok;
173
174
  metric.active = failed ? null : active.length;
174
175
  metric.last_line = line;
@@ -194,7 +195,7 @@ export async function runStartupChecks(db, { prefix = COLLECTION_PREFIX, log } =
194
195
  const owl = await owlStartupCheck(db, prefix);
195
196
  line = owl.line;
196
197
  } catch {
197
- line = `[emet] OWL RETIREMENT CHECK FAILED account=${OWL_ACCOUNT} active=unknown rung=R6 pointer=/owner/chooser`;
198
+ line = owlRetirementFailureLine();
198
199
  }
199
200
  write(line);
200
201
  return { line, exited: false };
@@ -3,7 +3,7 @@
3
3
  * board board 120, design the staging note; the practice is
4
4
  * docs/GUARDRAILS-AND-MONORAILS.md).
5
5
  *
6
- * Alex: "guardrails give a field of movement and execution to allow an AI to do
6
+ * the owner: "guardrails give a field of movement and execution to allow an AI to do
7
7
  * processes where rails guide the AI through a process flow without it
8
8
  * processing" (semantic ref 297). The floor, the character and the served
9
9
  * instructions are the guardrails. This file is the monorail - the car cannot
@@ -72,14 +72,14 @@ export const BOARD_DOC = 'STATE.md';
72
72
 
73
73
  /**
74
74
  * Per-bot close records (C1, 2026-10-01; review note "C1: per-bot close
75
- * records", Alex's decisions 1-7). The rule, stated once: a session written
75
+ * records", the owner's decisions 1-7). The rule, stated once: a session written
76
76
  * under a bot's source tag closes against that bot's own records
77
77
  * (bots/<tag>/HANDOFF.md); only that tag's own revisions in the session window
78
78
  * count; a bot tag never writes or cites the main HANDOFF.md or STATE.md.
79
79
  *
80
80
  * A tag T is a BOT TAG when bots/T/HANDOFF.md is listed in EMET_RECORD_DOCS,
81
81
  * the list that already marks ids as supersede-only records (tools.js
82
- * recordDocIds). No new setting: Alex adds one id per bot, and removing it
82
+ * recordDocIds). No new setting: the owner adds one id per bot, and removing it
83
83
  * puts that tag back on the host rules. Bot records are named by the source
84
84
  * tag itself (decision 1), so no tag-to-name mapping exists. These are pure
85
85
  * reads of settings - no database - so the graph can key its bot-handoff gate
@@ -361,7 +361,7 @@ export async function isSessionClosed(base, dbManager) {
361
361
  }
362
362
 
363
363
  /**
364
- * The closing exchange (2026-10-01, Alex's decision). In the closing turn the
364
+ * The closing exchange (2026-10-01, the owner's decision). In the closing turn the
365
365
  * assistant writes the handoff, gets `complete` from emet_session_close, and
366
366
  * only then replies - so the append of that last exchange always arrives
367
367
  * after the close. A closed session therefore takes exactly ONE more append:
@@ -786,7 +786,7 @@ export async function graphBefore(tool, args, dbManager, now = new Date()) {
786
786
  }
787
787
 
788
788
  // ---------------------------------------------------------------------------
789
- // Lab results (2026-09-29, Alex: "a 'lab results' location for storing the
789
+ // Lab results (2026-09-29, the owner: "a 'lab results' location for storing the
790
790
  // results in real time with time stamps so we can have a resource to look at
791
791
  // after the probationary period"). Every check the monorail makes - ok,
792
792
  // refuse, unknown (would_refuse in rows from before enforce-only) - is one timestamped row in
package/src/session.js CHANGED
@@ -355,7 +355,7 @@ export async function corroborateCoverage(db, baseId) {
355
355
  * The store check above answers "do the artifacts exist". It could not answer
356
356
  * "did the host actually do each step", and a host that skipped a step, or
357
357
  * never read the transcript back, could still walk away with `complete`
358
- * because the store happened to look fine. Alex's requirement (procedural ref 388,
358
+ * because the store happened to look fine. the owner's requirement (procedural ref 388,
359
359
  * 389, 391): the wrap steps HAVE TO BE FOLLOWED, whichever model is hosting,
360
360
  * and a skipped step blocks the close the same way a skipped opening is caught
361
361
  * at initialize.
@@ -975,7 +975,7 @@ export async function sessionClose(options = {}) {
975
975
  }
976
976
 
977
977
  // Store-level gaps make the session incomplete. Receipt problems, and a
978
- // transcript nothing independently corroborates, BLOCK it (Alex 2026-09-26,
978
+ // transcript nothing independently corroborates, BLOCK it (the owner 2026-09-26,
979
979
  // procedural ref 391: receipts required from the first deploy; the former
980
980
  // `unconfirmed` state is blocked, not passed).
981
981
  const storeMissing = [...missing];
package/src/setup.js CHANGED
@@ -250,7 +250,7 @@ export function registeredSourceTags() {
250
250
  return env.length ? [...new Set([...env, ...approvedTags()])] : env;
251
251
  }
252
252
 
253
- /** Attribution v2: the caller's real tag, not the server-wide default (Alex 2026-10-12 (c)).
253
+ /** Attribution v2: the caller's real tag, not the server-wide default (the owner 2026-10-12 (c)).
254
254
  * R2 adds the credential: grant_id (the token family) and door. Stdio has no grant. */
255
255
  export function callerAttribution() {
256
256
  const d = declaredAttribution();
@@ -335,7 +335,7 @@ export function summarizeSeed(doc_id, text, today = new Date().toISOString().sli
335
335
  }
336
336
 
337
337
  /**
338
- * One line per seed for the initialize payload (Alex, 2026-09-27, code audit
338
+ * One line per seed for the initialize payload (the owner, 2026-09-27, code audit
339
339
  * 2.3: "seeds can be one line since follow-up inquiry by the user will expand
340
340
  * it"). Before this every trigger's full test text was served - 13K characters
341
341
  * for 21 seeds with none due. The line keeps what a startup scan needs: the
@@ -876,7 +876,7 @@ export async function initialize(options = {}, hooks = {}) {
876
876
  if (late && late.length && typeof hooks.onLate === 'function') { try { hooks.onLate(late); } catch { /* recording never fails a page */ } }
877
877
  return late;
878
878
  };
879
- // Paged startup (2026-10-03, Alex approved paging; src/startup-pages.js).
879
+ // Paged startup (2026-10-03, the owner approved paging; src/startup-pages.js).
880
880
  // The full payload is built exactly as before; only its delivery changes.
881
881
  const raw = options.page;
882
882
  const wantAll = typeof raw === 'string' && raw.trim().toLowerCase() === 'all';
@@ -1130,9 +1130,9 @@ async function buildStartup(options = {}) {
1130
1130
  lessonsPending = { assessed: false, error: e && e.message ? e.message : String(e), pending: [] };
1131
1131
  }
1132
1132
 
1133
- // Stage 1 of the ref 358 startup amendment (2026-10-02; Alex's decision 2,
1133
+ // Stage 1 of the ref 358 startup amendment (2026-10-02; the owner's decision 2,
1134
1134
  // procedural ref 407: "emet_initialize is the startup doorway every bot
1135
- // calls; only first-time setup stays Alex's"). Initialize no longer creates
1135
+ // calls; only first-time setup stays the owner's"). Initialize no longer creates
1136
1136
  // the owner person: that write moved to setupComplete, the owner's
1137
1137
  // first-time setup, so initialize is read-only in every branch and matches
1138
1138
  // its annotation (annotations.js, READ) for every caller. A store with no
@@ -1168,7 +1168,7 @@ async function buildStartup(options = {}) {
1168
1168
  // and connection; the user's name does the animating.
1169
1169
  emet_line: EMET_LINE,
1170
1170
  emet_line_if_unanswered: EMET_LINE_UNANSWERED,
1171
- // 2026-09-26 Alex: prune may not drop the close-order tool. These two
1171
+ // 2026-09-26 the owner: prune may not drop the close-order tool. These two
1172
1172
  // fields MUST precede identity/character/user so a 20k host truncate
1173
1173
  // still receives them. The charter text may stay pointed; the steps may not.
1174
1174
  tools_in_force: {
@@ -1181,7 +1181,7 @@ async function buildStartup(options = {}) {
1181
1181
  'source tag on every write'
1182
1182
  ],
1183
1183
  on_wrap: ['working documents', 'episodic session record', 'board', 'verbatim transcript', 'HANDOFF last one write'],
1184
- // 2026-09-26 Alex (procedural ref 391): close is blocked without receipts.
1184
+ // 2026-09-26 the owner (procedural ref 391): close is blocked without receipts.
1185
1185
  confirm: 'emet_session_close with receipts for all four items and the handoff; say its speak lines in order; saved only when state is complete'
1186
1186
  },
1187
1187
  // Code audit 2026-09-27 (2.1): the short instructions ride up front with
@@ -1208,7 +1208,7 @@ async function buildStartup(options = {}) {
1208
1208
  ...retirementFields(installNotes)
1209
1209
  } : null,
1210
1210
  install_notes_doc_id: installDoc(),
1211
- // One line each since 2026-09-27 (Alex); the full seed is one read_doc away.
1211
+ // One line each since 2026-09-27 (the owner); the full seed is one read_doc away.
1212
1212
  seeds: seeds.map((s) => seedLine(s)),
1213
1213
  seeds_due: seeds.filter((s) => s.due && s.retired !== true).map((s) => s.doc_id),
1214
1214
  seeds_evaluated_on: today,
@@ -1354,8 +1354,8 @@ export async function setupComplete(answers = {}, logger = null, options = {}) {
1354
1354
  // guards (2026-10-02, Reviewer's stage 1 admission, conditions S1-S2):
1355
1355
  // - refused under a configured bot tag. The tag is self-declared and
1356
1356
  // emet_initialize takes no source until stage 2, so this is inert until a
1357
- // caller declares a bot tag that Alex has configured (stage 3);
1358
- // - refused on a store that is already set up unless `reissue: true` (Alex's
1357
+ // caller declares a bot tag that the owner has configured (stage 3);
1358
+ // - refused on a store that is already set up unless `reissue: true` (the owner's
1359
1359
  // answer 5): a re-run replaces the user's identity documents with new
1360
1360
  // versions, so it needs the owner's explicit word. Nothing is written
1361
1361
  // when either refusal fires.
package/src/tools.js CHANGED
@@ -798,7 +798,7 @@ export async function saveMemory(dbManager, content, layer = null, metadata = {}
798
798
  chain_sha256: chainSha256
799
799
  };
800
800
 
801
- // Option A correction back-pointer (Alex, 2026-10-02, procedural ref 405 item 4): a new entry may name the
801
+ // Option A correction back-pointer (the owner, 2026-10-02, procedural ref 405 item 4): a new entry may name the
802
802
  // entry it partly corrects. The parent must exist in the same layer, and is checked BEFORE the insert so a
803
803
  // refused correction writes nothing.
804
804
  if (validatedMetadata.corrects !== undefined && validatedMetadata.corrects !== null) {
@@ -1015,7 +1015,7 @@ export async function resolveSupersessionChain(read, entry, maxHops = 50) {
1015
1015
  }
1016
1016
 
1017
1017
  /**
1018
- * Option A correction back-pointer (Alex, 2026-10-02, procedural ref 405 item 4; Reviewer's sketch in his
1018
+ * Option A correction back-pointer (the owner, 2026-10-02, procedural ref 405 item 4; Reviewer's sketch in his
1019
1019
  * second read of episodic ref 604). Episodic entries are immutable events, so a PARTIAL correction cannot
1020
1020
  * supersede its parent: supersession would put an end date on the whole parent and present every item
1021
1021
  * in it as replaced. Instead the new entry carries metadata.corrects = <parent id>, and the parent is MARKED:
@@ -2758,7 +2758,7 @@ export async function readDocHistory(doc_id, version = null, logger = null) {
2758
2758
  }
2759
2759
 
2760
2760
  /**
2761
- * Records are superseded, never amended (2026-09-26, Alex: "enforce
2761
+ * Records are superseded, never amended (2026-09-26, the owner: "enforce
2762
2762
  * supersession ISO practices"). ISO 9001 7.5.3 / ISO 13485 4.2.5 separate
2763
2763
  * DOCUMENTS, which are revised under control, from RECORDS, which are evidence
2764
2764
  * of what happened and are not altered. The handoff is a record: it attests to
@@ -2805,7 +2805,7 @@ function _refuseRecordAmendment(doc_id, tool) {
2805
2805
  * stamps as updated_by: the caller's, else EMET_SOURCE_TAG. A records rule like
2806
2806
  * the record-amendment refusal above, so it applies in every graph mode.
2807
2807
  *
2808
- * Two further guards on the same doors (2026-10-01, Alex's 4:17 PM approval to
2808
+ * Two further guards on the same doors (2026-10-01, the owner's 4:17 PM approval to
2809
2809
  * make the bot tags fully functional; Reviewer's review of PR 3):
2810
2810
  * - REGISTERED HOSTS ONLY on the main records. With a source registry
2811
2811
  * configured (EMET_SOURCE_TAGS, parsed as save_to_layer parses it), the
@@ -3005,18 +3005,13 @@ function refuseBotMainRecordOne(doc_id, updated_by, tool, options = {}) {
3005
3005
  }
3006
3006
 
3007
3007
  /**
3008
- * retire_doc and rename_doc are not bot tools at all (Alex, 2026-10-01 4:27 PM
3009
- * ET, decision 1, as relayed: "retire_doc isn't needed for bots: keep it
3010
- * Alex/hub only and out of every bot profile"; procedural ref 407; stage 1 of the
3011
- * ref 358 startup amendment, admitted by Reviewer 2026-10-02). A profile is not a
3012
- * security boundary, so the door refuses them under ANY bot tag, for every
3013
- * doc_id - not only the main records refuseBotMainRecord covers. ref 407(1)
3014
- * covers retire_doc only; rename_doc stays Alex-only under procedural ref 409
3015
- * (Alex-confirmed 2026-10-02 7:50: every other Alex-only tool is
3016
- * unchanged), and it retires its old id internally, so it is refused too. The tag is the
3017
- * one the door would stamp: the caller's, else EMET_SOURCE_TAG. Hosts and the
3018
- * hub have no bots/<tag>/HANDOFF.md record and are unaffected. Nothing is
3019
- * written when refused.
3008
+ * retire_doc and rename_doc are not bot tools. They stay with the owner and
3009
+ * the hub. A profile is not a security boundary, so the door refuses them
3010
+ * under any bot tag, for every document, not only the main records
3011
+ * refuseBotMainRecord covers. Rename retires its old id, so it is refused
3012
+ * too. The tag is the one the door would stamp: the caller's, else
3013
+ * EMET_SOURCE_TAG. Hosts and the hub have no bots/<tag>/HANDOFF.md record
3014
+ * and are unaffected. Nothing is written when refused.
3020
3015
  */
3021
3016
  export function refuseBotRetireRename(tag, tool) {
3022
3017
  for (const t of rightsTags(tag)) refuseBotRetireRenameOne(t, tool);
@@ -3035,7 +3030,7 @@ function refuseBotRetireRenameOne(tag, tool) {
3035
3030
  throw new ValidationError(
3036
3031
  'source',
3037
3032
  `${tool} is not a bot tool: ${t} is a bot tag, and ` +
3038
- `renaming documents stays the owner-only (procedural ref 409)` +
3033
+ `renaming documents stays with the owner and the hub` +
3039
3034
  `. Ask the hub. Nothing was changed.`
3040
3035
  , { code: 'EMET-MEM-020' });
3041
3036
  }
@@ -4014,7 +4009,7 @@ export async function memberAccessCreate(dbManager, { display_name, days, source
4014
4009
  const who = currentActor();
4015
4010
  const aiWriter = isOwlIdentity(who) || (who.transport === 'http' && (isMasterBlasterActor(who) || actorIsCos(who) || (who.sourceTag && isBotTag(who.sourceTag))));
4016
4011
  if (aiWriter) {
4017
- throw new ValidationError('An AI cannot grant a writer. Alex decides on /owner/chooser with an owner session. Nothing was created.', { code: 'EMET-AUTH-043' });
4012
+ throw new ValidationError('An AI cannot grant a writer. The owner decides on /owner/chooser with an owner session. Nothing was created.', { code: 'EMET-AUTH-043' });
4018
4013
  }
4019
4014
  const { registeredSourceTags } = await import('./setup.js');
4020
4015
  const actor = await codeCreator(dbManager);
@@ -4603,7 +4598,7 @@ const TOOL_DEFINITIONS = [
4603
4598
  },
4604
4599
  {
4605
4600
  name: "emet_client_approve",
4606
- description: "Approve or deny a pending client tag, or prepare a new writer. decision prepare drafts a writer request: client, claimed tag, door (member or guest), reason, and correlation_id. The hub login prepares that draft. The server validates it. Validation is a check and is not approval. Alex decides a new writer on /owner/chooser with the owner session. No AI grants a writer, including the hub login and the retired owl account. The owner login may still approve or deny a pending client tag. Pending client tags are listed in emet_initialize and emet_status as client_approvals.",
4601
+ description: "Approve or deny a pending client tag, or prepare a new writer. decision prepare drafts a writer request: client, claimed tag, door (member or guest), reason, and correlation_id. The hub login prepares that draft. The server validates it. Validation is a check and is not approval. The owner decides a new writer on /owner/chooser with the owner session. No AI grants a writer, including the hub login and the retired owl account. The owner login may still approve or deny a pending client tag. Pending client tags are listed in emet_initialize and emet_status as client_approvals.",
4607
4602
  inputSchema: {
4608
4603
  type: "object",
4609
4604
  properties: {
@@ -5,7 +5,7 @@ import { setting } from './env.js';
5
5
  export const EXCHANGES_PER_PART = 8;
6
6
 
7
7
  /**
8
- * The zone a session id is dated in (2026-10-01, Alex's approval; dev team plan
8
+ * The zone a session id is dated in (2026-10-01, the owner's approval; dev team plan
9
9
  * Part B, fix 1). Ids were dated from UTC, so an evening session in New York
10
10
  * carried the next day's date (episodic ref 600: `2026-10-01_grok-init` minted on
11
11
  * the night of 9/30). The day is now the user's local day: EMET_TIMEZONE, an
@@ -86,7 +86,7 @@ export function shouldStartNewPart(exchangeCount) {
86
86
  }
87
87
 
88
88
  /**
89
- * A closed session is never reopened (2026-10-01, Alex's approval; episodic
89
+ * A closed session is never reopened (2026-10-01, the owner's approval; episodic
90
90
  * 598: a session_open returned the previous night's CLOSED id and the next
91
91
  * append landed in it as exchange 9). Returns the first of base, base-2,
92
92
  * base-3, ... that is free, and the ids it passed over with the reason.
package/src/version.js CHANGED
@@ -1,2 +1,2 @@
1
1
  /** The server's version, in one place. tests/version.test.js keeps it equal to package.json. */
2
- export const EMET_VERSION = '0.3.5';
2
+ export const EMET_VERSION = '0.3.6';
@@ -2,7 +2,7 @@
2
2
 
3
3
  **Template control:** id `templates/SESSIONDISPOSITION.md` · revision: the store's version governs · status: IN FORCE · approved: user · 2026-10-28 · session-disposition brief · owner: EMET (base) · built to `templates/STANDARD.md` · supersedes: none
4
4
  **Fills:** `records/<name>.md` — one session, or a batch of sessions, noted under document control. **Every `records/*` document is validated against this template by its id** (`templates/STANDARD.md` §9 map). Leaving out the control block does not exempt it (§9a L4).
5
- **Failure it prevents:** `emet_session_close` compares receipts with the store as it stands that day, so a close attempted days later blocks, and a third blocked close escalates. A fabricated receipt is not a close. Reference case, 2026-10-28: an early `save_transcript` for `2026-10-28_mb-main-chat-backfill-2026-10-26`, inserted_id `6ae1df09c4da08b2cadc1fb4`, eight exchanges, first assistant line opening `grok-bot-vm-453484571 Wed Oct 28 07:38:40 EDT 2026`. That row is a duplicate of `2026-10-28_mb-main-chat-backfill-2026-10-26_verbatim_part1`, written by `emet_transcript_append`, machine line `no local shell MCP`. The duplicate is marked SUPERSEDED and retained. It is not deleted, and it is not a close.
5
+ **Failure it prevents:** `emet_session_close` compares receipts with the store as it stands that day, so a close attempted days later blocks, and a third blocked close escalates. A fabricated receipt is not a close. Reference case, 2026-10-28: an early `save_transcript` for `2026-10-28_mb-main-chat-backfill-2026-10-26`, inserted_id `6ae1df09c4da08b2cadc1fb4`, eight exchanges, first assistant line opening `example-host Wed Oct 28 07:38:40 EDT 2026`. That row is a duplicate of `2026-10-28_mb-main-chat-backfill-2026-10-26_verbatim_part1`, written by `emet_transcript_append`, machine line `no local shell MCP`. The duplicate is marked SUPERSEDED and retained. It is not deleted, and it is not a close.
6
6
  **Completion rules:** `templates/STANDARD.md` §5, unmodified.
7
7
 
8
8
  ---
@@ -27,7 +27,7 @@ The early `save_transcript` stays SUPERSEDED. It is a duplicate of the Oct 26 ve
27
27
 
28
28
  | session_id | opened_at | last_saved_at | disposition | reason | evidence | carried_forward_to | inserted_id |
29
29
  |---|---|---|---|---|---|---|---|
30
- | 2026-10-28_mb-main-chat-backfill-2026-10-26 | None — this row dispositions the early save_transcript, not the session open. | None — the live Oct 26 backfill is the verbatim part, not this inserted id. | SUPERSEDED_BY_LATER_SESSION | Early save_transcript is a SUPERSEDED duplicate of the verbatim_part1 append path. The transcript row is retained and is not deleted. | inserted_id 6ae1df09c4da08b2cadc1fb4 has 8 exchanges. The first assistant message opens with grok-bot-vm-453484571 Wed Oct 28 07:38:40 EDT 2026. The correct Oct 26 backfill is 2026-10-28_mb-main-chat-backfill-2026-10-26_verbatim_part1 via emet_transcript_append, machine line no local shell MCP. | 2026-10-28_mb-main-chat-backfill-2026-10-26_verbatim_part1 | 6ae1df09c4da08b2cadc1fb4 |
30
+ | 2026-10-28_mb-main-chat-backfill-2026-10-26 | None — this row dispositions the early save_transcript, not the session open. | None — the live Oct 26 backfill is the verbatim part, not this inserted id. | SUPERSEDED_BY_LATER_SESSION | Early save_transcript is a SUPERSEDED duplicate of the verbatim_part1 append path. The transcript row is retained and is not deleted. | inserted_id 6ae1df09c4da08b2cadc1fb4 has 8 exchanges. The first assistant message opens with example-host Wed Oct 28 07:38:40 EDT 2026. The correct Oct 26 backfill is 2026-10-28_mb-main-chat-backfill-2026-10-26_verbatim_part1 via emet_transcript_append, machine line no local shell MCP. | 2026-10-28_mb-main-chat-backfill-2026-10-26_verbatim_part1 | 6ae1df09c4da08b2cadc1fb4 |
31
31
  | 2026-10-28_mb-main-chat-backfill-2026-10-26_verbatim_part1 | None — not an overdue-inventory timestamp. This row records the open append path. | None — this row does not invent a save time. The append path holds the substance. | RETAINED_OPEN_CRITICAL | This open backfill still needs a formal emet_session_close. The filtered 31-exchange append path is the complete substance for Oct 25 through Oct 27. Do not invent a close or a receipt. | Coverage note 2026-10-28. Late Oct 25 PM and Oct 27 after about 8:40 had no owner main-chat turns (hidden wakes only). | None — formal close is still owed on this session. | None — this row dispositions the session, not one transcript document. |
32
32
 
33
33
  ---
@@ -5,7 +5,7 @@
5
5
  * Licensed under the MIT License - see LICENSE
6
6
  *
7
7
  * Load gate for a client RUNE. The server never runs the file. This module
8
- * is the scanner EMET-GEMs vendors: keep it byte-identical, including the
8
+ * is the scanner a library checkout vendors: keep it byte-identical, including the
9
9
  * corpus under verification/load-gate/. acorn 8.19.0 only. A parse error
10
10
  * fails closed. There is no second scan and no child process.
11
11
  *