@zenera/cli 1.1.16 → 1.1.18

package/README.md

@@ -125,7 +125,7 @@ In the editor, describe the job in `SPECIFICATION.md`, then send this in the
 agent chat:
 
 ```
-/sync-with-spec
+/spec-sync-project
 ```
 
 It creates the specialists, responsibilities, tool access and handoffs the job
@@ -145,7 +145,7 @@ zen run my-project              # no prompt given: opens a full-screen terminal
 zen check my-project            # validate the project and every file it names
 zen inspect                     # open the last run's report.html
 zen list --sessions             # every project, its sessions and last run
-echo "triage this" | zen run my-project --quiet | jq
+echo "triage this" | zen run my-project --json | jq
 ```
 
 Standing inside the project, the name is optional: a bare `zen run`, `zen check`
@@ -156,7 +156,7 @@ session and uses the current directory as the workspace, with write access.
 `--session`, `--workspace` and `--read-only` override that.
 
 To change what the system does, update `SPECIFICATION.md` and send
-`/sync-with-spec` again. The next section explains that workflow in detail.
+`/spec-sync-project` again. The next section explains that workflow in detail.
 
 ## You write the specification; a coding agent writes the system
 
@@ -202,12 +202,15 @@ In the window that opens:
 
 1. **Edit `SPECIFICATION.md`** - what the system is for, which agents exist,
    what each may reach for, and what _done_ means.
-2. **Open the chat panel and send `/sync-with-spec`.** It reads the
+2. **Open the chat panel and send `/spec-sync-project`.** It reads the
    specification and every file implementing it, works out the difference in
    both directions, and changes the smallest thing that closes each gap.
-3. **Read `SPECIFICATION-FEEDBACK.md`**, which it writes for anything it could
-   not do without guessing. Answer its questions by editing `SPECIFICATION.md`
-    - not by editing prompts - and send `/sync-with-spec` again.
+3. **Answer `SPECIFICATION-FEEDBACK.md`**, which it writes for anything it could
+   not do without guessing. Every question carries two to four written-out
+   answers, each with a `[ ]` box: tick one, delete the ones you do not want, or
+   write your own in their place.
+4. **Send `/spec-apply-feedback`.** It folds the answers you chose into
+   `SPECIFICATION.md` - never into a prompt - and then `/spec-sync-project` again.
 
 Back in the terminal: `zen check`, then `zen run`.
 
@@ -218,15 +221,16 @@ They are prompt files under `.github/prompts/`, which VS Code and its forks
 offer as chat slash-commands. Both `zen init` and `zen open` write them fresh,
 so they never go stale - and edits to them do not survive.
 
-| In chat           | Does                                                        |
-| ----------------- | ----------------------------------------------------------- |
-| `/sync-with-spec` | Makes every file match `SPECIFICATION.md`, both directions. |
-| `/review-project` | Reads the project as a reviewer would, and reports.         |
-| `/new-agent`      | Adds an agent - prompt, wiring and hand-offs.               |
-| `/new-skill`      | Adds a skill under `agents/skills/`.                        |
+| In chat                | Does                                                             |
+| ---------------------- | ---------------------------------------------------------------- |
+| `/spec-sync-project`   | Makes every file match `SPECIFICATION.md`, both directions.      |
+| `/spec-apply-feedback` | Folds your answers in `SPECIFICATION-FEEDBACK.md` into the spec. |
+| `/project-review`      | Reads the project as a reviewer would, and reports.              |
+| `/new-agent`           | Adds an agent - prompt, wiring and hand-offs.                    |
+| `/new-skill`           | Adds a skill under `agents/skills/`.                             |
 
 In an editor that does not support prompt files, paste the contents of
-`.github/prompts/sync-with-spec.prompt.md` into its chat instead - it is only a
+`.github/prompts/spec-sync-project.prompt.md` into its chat instead - it is only a
 prompt. Alongside them, `.github/copilot-instructions.md` is the standing brief
 that explains this runtime to whatever agent is reading, and
 `.github/skills/zen-cli/` is the CLI's own reference for it.
@@ -236,7 +240,7 @@ that explains this runtime to whatever agent is reading, and
 ### The self-improving loop
 
 ```
-edit SPECIFICATION.md → /sync-with-spec → zen check → zen run
+edit SPECIFICATION.md → /spec-sync-project → zen check → zen run
         ↑                                                  ↓
         └───── refine the spec <-──── zen inspect <-───────┘
 ```
@@ -295,7 +299,7 @@ Read a codebase and leave a short written note about it in the workspace.
 - Every claim in it comes from a file that was actually read.
 ```
 
-Send `/sync-with-spec` in the editor's chat. It writes `agents.yaml` - who
+Send `/spec-sync-project` in the editor's chat. It writes `agents.yaml` - who
 exists, and what each may reach for:
 
 ```yaml
@@ -378,20 +382,24 @@ is a command on their machine too.
 Prompts, agent wiring, skills and tool selections are all files - Markdown and
 YAML - rather than code buried inside an application. So the folder can be
 committed, copied to another machine, reviewed in a pull request, and handed to
-someone else who runs it with one command. Credentials live in `$HOME`, never in
-the project, so there is no secret to strip before sharing it.
+someone else who runs it with one command. Model credentials live in `$HOME`,
+never in the project, and anything the project itself needs goes in a
+git-ignored `.env` - so there is no secret to strip before sharing it.
 
 ```
 my-project/
     SPECIFICATION.md             what this is for - the intent
-    INSTRUCTIONS.md              house rules, prepended to every agent
     agents.yaml                  who exists, what they may reach for
     agents/
+        instructions.md          house rules, prepended to every agent
+        <topic>-instructions.md  more of them, one subject at a time
         prompts/<name>.md        each agent's own brief
         skills/<name>/           knowledge loaded on demand, not always-on
     assets/                      reference material, read-only at /assets
     sandbox/Dockerfile           the container commands run in
     scripts/_setup.sh            the one command that initialises the project
+    .spec-sync/                  what the last sync with the spec applied
+    .env                         this project's environment - git-ignored
     sessions/                    one workspace, memory and trajectory each
         <id>/
             workspace/           what the agents can read and write
@@ -457,6 +465,7 @@ remove resources beyond the current project.
 | `key`     | The credential keyring - add, check, switch, remove.                 |
 | `models`  | What this machine can use - list, search, test, pick.                |
 | `run`     | Runs the project - the TUI on a terminal, a single answer otherwise. |
+| `meta`    | Runs the meta agent over the project - the one that edits it.        |
 | `inspect` | Opens or rebuilds a run's `report.html`.                             |
 | `memory`  | What the agents remember - size, listing, one node, or a whole page. |
 | `check`   | Validates the project and every file it names, and asks the models.  |
@@ -531,6 +540,88 @@ an accident.
 | `4`       | no usable credential |
 | `5`       | sandbox unavailable  |
 
+## The meta agent - the agent that works on the project
+
+`zen run` runs the agents your project describes. `zen meta` runs an agent _over_
+the project: a coding agent rooted at the project directory, which reads
+`SPECIFICATION.md`, changes `agents.yaml`, the prompts and the skills, and runs
+whatever commands it needs to check itself - on the keys already in your
+keyring, and not on anybody's subscription. It is the loop your editor's chat
+panel drives, as a command you can put in a script.
+
+```sh
+zen meta run "what does this project do?"
+zen meta run acme "review the last commit"
+zen meta prompts                             # which /<name> prompts this project has
+zen meta run /project-review                 # .github/prompts/project-review.prompt.md
+zen meta run acme /spec-sync-project agents/triage.md
+git diff | zen meta run "what broke?" --allow-tool read
+zen meta run --dry-run "hello"               # what would run, secrets masked
+```
+
+Every prompt goes through `run`, whether you typed it or it is a file: the
+alternative was a verb that appeared only sometimes, and a bare word after
+`zen meta` that was a subcommand on Tuesdays and a question on Wednesdays.
+
+It uses every tool without asking. The prompts it runs were written for an
+editor, where reading files, writing them and running commands is the whole
+point, and a `-p` run has nobody sitting there to answer the question - so the
+alternative is not a safer run but a stalled one. `--allow-tool read`,
+`--allow-tool 'shell(git:*)'` narrow it to what a particular run needs, and
+`--ask` puts the question back. It is your own checkout and your own key either
+way, so treat it as you would `git push`: fine on a project you own, worth
+narrowing on one you do not.
+
+The answer goes to stdout and the progress to stderr, like everything else here,
+so `zen meta run "…" > out.md` keeps the answer alone.
+
+Progress is a small window that rewrites itself rather than a transcript
+scrolling past, because only the last few steps say what it is doing now. The
+whole of it - the prompt, every tool call, the answer - is written as it happens
+to `.tmp/logs/meta.<when>.log`, whose path is printed before the run starts, so
+`tail -f` in another terminal follows a run already going. The answer alone is
+kept beside it as `.tmp/logs/meta.<when>.md`, named again once it has printed -
+a terminal turns that into something you click.
+
+`zen meta run /<name>` is the part an editor cannot do for you from a script.
+The prompts under `.github/prompts/` - `/spec-sync-project`, `/project-review` -
+are what a chat panel offers as slash commands, and nothing outside an editor
+reads them. zen reads one, drops the frontmatter and sends the body - so the
+same prompt runs from the editor, from a terminal and from CI.
+
+A terminal has no menu dropping down as you type a slash, so
+`zen meta prompts` lists what this project holds, each with its description.
+`zen meta run` with nothing to say asks the same question interactively.
+
+### Which model it runs on
+
+```sh
+zen meta model                            # the whole chain, with the winner marked
+zen meta model vertex/gemini-3.8-flash    # for every project on this machine
+zen meta model --local openai/gpt-5.6-sol # for this project only, via its .env
+zen meta model --pick                     # choose from providers you hold a key for
+```
+
+Highest first: `--model`, `ZENERA_META_MODEL` in the shell, the same variable in
+the project's `.env`, `zen meta model`, then the project's `agents.yaml`
+`model:`. When none of them answers, a key does: the best model zen knows of
+for the first provider you hold one for, named on stderr as it starts, so a
+fresh machine needs a key and nothing else.
+The store is separate from `agents.yaml` on purpose - the model a coding agent
+runs on is a personal choice about a tool, like a key, where `agents.yaml` is a
+committed decision about the project's own agents.
+
+Not every model works. The meta agent declares its tools in a form only the
+reasoning APIs accept, so use a reasoning model - `gpt-5.6-sol`,
+`claude-opus-5`, `gemini-3.8-flash`. Those three are also what it falls back
+to, and what `zen meta model --pick` offers.
+
+Setting one asks it a one-word question before storing it, so a model that will
+not answer is refused with the provider's own words rather than at the start of
+the next run. A typed prefix that names no provider - `vertes/...` - is caught
+before that, because it would otherwise be sent whole to a vendor you never
+named. `--force` stores a ref without asking.
+
 ## Credentials
 
 One keyring serves every provider, and a key goes in the same way whatever it

package/dist/banner.d.ts

@@ -1,3 +1,19 @@
+/** One per sub-brand, so two of them are told apart before the word is read. */
+declare const ACCENTS: {
+    readonly orange: {
+        readonly face: "\u001B[1;38;5;208m";
+        readonly shade: "\u001B[0;38;5;130m";
+    };
+    readonly cyan: {
+        readonly face: "\u001B[1;38;5;51m";
+        readonly shade: "\u001B[0;38;5;37m";
+    };
+    readonly mint: {
+        readonly face: "\u001B[1;38;5;121m";
+        readonly shade: "\u001B[0;38;5;78m";
+    };
+};
+export type Hue = keyof typeof ACCENTS;
 export interface BannerText {
     /** drawn white */
     head: string;
@@ -5,6 +21,8 @@ export interface BannerText {
     accent: string;
     /** the line underneath, dim */
     subtitle: string;
+    /** which neon to use. Default: orange, the `zen` mark. */
+    hue?: Hue;
 }
 export interface PrintBannerOptions {
     /** Render the block face even when the terminal is narrower than it. */
@@ -14,4 +32,5 @@ export declare const NEO_BANNER: BannerText;
 export declare function bannerLines(text: BannerText, columns?: number): string[];
 /** Narration, and only for someone watching. */
 export declare function printBanner(text: BannerText, options?: PrintBannerOptions): void;
+export {};
 //# sourceMappingURL=banner.d.ts.map

package/dist/banner.js

@@ -1,5 +1,4 @@
-import { styleText } from 'node:util';
-import { dim, note } from "./term.js";
+import { dim, note, styled } from "./term.js";
 // ---------------------------------------------------------------------------
 // The banner
 //
@@ -65,11 +64,14 @@ function big(word) {
 }
 /** A wordmark rather than a sentence, so the letters are set apart. */
 const spaced = (s) => [...s.toUpperCase()].join(' ');
-/** Whether anything should be coloured at all — `styleText`'s own answer. */
-const styling = () => styleText('dim', '.') !== '.';
 const RESET = '\u001b[0m';
 const HEAD = { face: '\u001b[1;38;5;231m', shade: '\u001b[0;38;5;244m' };
-const ACCENT = { face: '\u001b[1;38;5;208m', shade: '\u001b[0;38;5;130m' };
+/** One per sub-brand, so two of them are told apart before the word is read. */
+const ACCENTS = {
+    orange: { face: '\u001b[1;38;5;208m', shade: '\u001b[0;38;5;130m' },
+    cyan: { face: '\u001b[1;38;5;51m', shade: '\u001b[0;38;5;37m' },
+    mint: { face: '\u001b[1;38;5;121m', shade: '\u001b[0;38;5;78m' },
+};
 /** Runs of stroke and runs of bevel, alternating; spaces stay unstyled. */
 const RUNS = /[█▀▄]+|[^█▀▄ ]+/g;
 const isStroke = (run) => /[█▀▄]/.test(run);
@@ -77,11 +79,11 @@ const isStroke = (run) => /[█▀▄]/.test(run);
  * Two tones per word — without the darker bevel the letter is a flat outline
  * and the shadow reads as part of the stroke.
  */
-const paint = (text, tone) => styling()
+const paint = (text, tone) => styled()
     ? text.replace(RUNS, (run) => `${isStroke(run) ? tone.face : tone.shade}${run}${RESET}`)
     : text;
 /** The wordmark has no bevel to shade, so it is all face. */
-const flat = (text, tone) => styling() ? `${tone.face}${text}${RESET}` : text;
+const flat = (text, tone) => styled() ? `${tone.face}${text}${RESET}` : text;
 export const NEO_BANNER = {
     head: 'Zenera',
     accent: 'Neo',
@@ -92,17 +94,18 @@ export function bannerLines(text, columns = process.stderr.columns || 80) {
     // Trimmed: the widest accent row is the banner's right edge, and a pad left
     // inside a styled string cannot be trimmed away later.
     const accent = big(text.accent).map((row) => row.trimEnd());
+    const tone = ACCENTS[text.hue ?? 'orange'];
     const margin = ' '.repeat(INDENT);
     const width = INDENT + head[0].length + GAP + Math.max(...accent.map((row) => row.length));
     if (width > columns) {
         return [
-            `${margin}${flat(text.head.toUpperCase(), HEAD)} ${flat(text.accent.toUpperCase(), ACCENT)}`,
+            `${margin}${flat(text.head.toUpperCase(), HEAD)} ${flat(text.accent.toUpperCase(), tone)}`,
             `${margin}${dim(spaced(text.subtitle))}`,
         ];
     }
     const lines = [];
     for (let r = 0; r < HEIGHT; r++) {
-        lines.push(`${margin}${paint(head[r], HEAD)}${' '.repeat(GAP)}${paint(accent[r], ACCENT)}`);
+        lines.push(`${margin}${paint(head[r], HEAD)}${' '.repeat(GAP)}${paint(accent[r], tone)}`);
     }
     // Rules on both sides, centred under the art: the subtitle is a caption,
     // and without them a short line of spaced capitals floats.

package/dist/catalog.js

@@ -38,29 +38,28 @@ const CATALOG_VERSION = 1;
  */
 export const CURATED = {
     openai: [
-        { id: 'gpt-4o-mini', roles: ['chat'], contextLength: 128_000 },
-        { id: 'gpt-4o', roles: ['chat'], contextLength: 128_000 },
+        { id: 'gpt-5.6-luna', roles: ['chat'], contextLength: 1_050_000 },
+        { id: 'gpt-5.6-terra', roles: ['chat'], contextLength: 1_050_000 },
         { id: 'text-embedding-3-small', roles: ['embedding'], dimensions: 1536 },
         { id: 'text-embedding-3-large', roles: ['embedding'], dimensions: 3072 },
     ],
     anthropic: [
-        { id: 'claude-haiku-4-5', roles: ['chat'], contextLength: 200_000 },
-        { id: 'claude-sonnet-4-5', roles: ['chat'], contextLength: 200_000 },
+        { id: 'claude-sonnet-5', roles: ['chat'], contextLength: 1_000_000 },
+        { id: 'claude-opus-5', roles: ['chat'], contextLength: 1_000_000 },
     ],
     google: [
-        { id: 'gemini-2.5-flash', roles: ['chat'], contextLength: 1_048_576 },
-        { id: 'gemini-2.5-pro', roles: ['chat'], contextLength: 1_048_576 },
-        { id: 'gemini-embedding-001', roles: ['embedding'], dimensions: 3072 },
+        { id: 'gemini-3.5-flash-lite', roles: ['chat'], contextLength: 1_048_576 },
+        { id: 'gemini-3.8-flash', roles: ['chat'], contextLength: 1_048_576 },
+        { id: 'gemini-embedding-2', roles: ['embedding'] },
     ],
     vertex: [
-        { id: 'gemini-2.5-flash', roles: ['chat'], contextLength: 1_048_576 },
-        { id: 'gemini-2.5-pro', roles: ['chat'], contextLength: 1_048_576 },
-        { id: 'gemini-embedding-001', roles: ['embedding'], dimensions: 3072 },
-        { id: 'text-embedding-005', roles: ['embedding'], dimensions: 768 },
+        { id: 'gemini-3.5-flash-lite', roles: ['chat'], contextLength: 1_048_576 },
+        { id: 'gemini-3.8-flash', roles: ['chat'], contextLength: 1_048_576 },
+        { id: 'gemini-embedding-2', roles: ['embedding'] },
     ],
     openrouter: [
-        { id: 'openai/gpt-4o-mini', roles: ['chat'] },
-        { id: 'anthropic/claude-haiku-4.5', roles: ['chat'] },
+        { id: 'openai/gpt-5.6-luna', roles: ['chat'], contextLength: 1_050_000 },
+        { id: 'anthropic/claude-sonnet-5', roles: ['chat'], contextLength: 1_000_000 },
         { id: 'openai/text-embedding-3-small', roles: ['embedding'], dimensions: 1536 },
     ],
 };
@@ -73,23 +72,23 @@ export const CURATED = {
  */
 export const PREFERRED = {
     openai: {
-        chat: ['gpt-4o-mini', 'gpt-4o'],
+        chat: ['gpt-5.6-luna', 'gpt-5.6-terra'],
         embedding: ['text-embedding-3-small', 'text-embedding-3-large'],
     },
     anthropic: {
-        chat: ['claude-haiku-4-5', 'claude-sonnet-4-5'],
+        chat: ['claude-sonnet-5', 'claude-opus-5'],
         embedding: [],
     },
     google: {
-        chat: ['gemini-2.5-flash', 'gemini-2.5-pro'],
-        embedding: ['gemini-embedding-001'],
+        chat: ['gemini-3.5-flash-lite', 'gemini-3.5-flash'],
+        embedding: ['gemini-embedding-2'],
     },
     vertex: {
-        chat: ['gemini-2.5-flash', 'gemini-2.5-pro'],
-        embedding: ['text-embedding-005', 'gemini-embedding-001'],
+        chat: ['gemini-3.5-flash-lite', 'gemini-3.5-flash'],
+        embedding: ['gemini-embedding-2'],
     },
     openrouter: {
-        chat: ['openai/gpt-4o-mini', 'anthropic/claude-haiku-4.5'],
+        chat: ['openai/gpt-5.6-luna', 'anthropic/claude-sonnet-5'],
         embedding: ['openai/text-embedding-3-small'],
     },
 };

package/dist/command.d.ts

@@ -1,3 +1,4 @@
+import type { BannerText } from './banner.ts';
 export interface Context {
     readonly args: readonly string[];
     readonly json: boolean;
@@ -9,6 +10,8 @@ export interface Command {
     readonly usage: string;
     /** lines printed under the usage line by `zen help <command>` */
     readonly details?: readonly string[];
+    /** printed instead of the `zen` banner, so the sub-brand shows through */
+    readonly banner?: BannerText;
     /**
      * Answers `--help` itself, for a command whose arguments select a page the
      * frame cannot see. Without it the frame prints this command's own page.

package/dist/commands/check.js

@@ -1,6 +1,7 @@
 import { existsSync, statSync } from 'node:fs';
 import { basename, dirname, resolve } from 'node:path';
 import { one, parse } from "../args.js";
+import { loadProjectEnv } from "../env.js";
 import { KeyStore } from "../keys.js";
 import { duration } from "../narrate.js";
 import { Registry } from "../projects.js";
@@ -75,7 +76,10 @@ export const check = {
         const name = entry?.name ?? basename(dir);
         // Materialised first, so the credential verdicts are the ones a run
         // would reach: a key in the environment and a key in the keyring are
-        // the same key by the time the library asks.
+        // the same key by the time the library asks. The project's own `.env`
+        // goes first for the same reason — `zen run` reads it, so a check that
+        // did not would report a missing credential for a project that runs.
+        loadProjectEnv(dir);
         const keys = await KeyStore.open();
         keys.materialize();
         const bar = progress();

package/dist/commands/index.js

@@ -5,6 +5,7 @@ import { inspect } from "./inspect.js";
 import { key } from "./key.js";
 import { list } from "./list.js";
 import { memory } from "./memory.js";
+import { meta } from "./meta.js";
 import { models } from "./models.js";
 import { open } from "./open.js";
 import { run } from "./run.js";
@@ -19,6 +20,7 @@ export const COMMANDS = {
     init,
     list,
     run,
+    meta,
     open,
     key,
     models,
@@ -37,6 +39,7 @@ export const ALIASES = {
     model: 'models',
     validate: 'check',
     doctor: 'check',
+    copilot: 'meta',
     report: 'inspect',
     mem: 'memory',
     edit: 'open',
@@ -49,7 +52,7 @@ export const EXTERNAL = {
         summary: 'A mock API from an openapi/swagger document.',
         usage: 'zen faker <serve|build|cache> [spec...]',
         install: 'npm i -g @zenera/faker',
-        banner: { head: 'Zenera', accent: 'Faker', subtitle: 'Mock API Server' },
+        banner: { head: 'Zenera', accent: 'Faker', subtitle: 'Mock API Server', hue: 'mint' },
     },
     rag: {
         package: '@zenera/rag',

package/dist/commands/init.js

@@ -2,6 +2,7 @@ import { readProjectConfig } from '@zenera/neo';
 import { existsSync, readdirSync } from 'node:fs';
 import { basename, resolve } from 'node:path';
 import { one, parse } from "../args.js";
+import { PREFERRED } from "../catalog.js";
 import { ensureHome } from "../home.js";
 import { resolveBuild } from "../image.js";
 import { envNames, isProvider, keyId, KeyStore, PROVIDERS } from "../keys.js";
@@ -43,6 +44,36 @@ const DEFAULT_MODEL = {
     vertex: { ref: 'vertex:gemini-3.5-flash' },
     openrouter: { ref: 'openrouter:inclusionai/ling-3.0-flash-fin:free' },
 };
+/**
+ * What vectorises the project's memory, given the provider it ends up on.
+ *
+ * Taken from the same table `zen models pick` ranks by rather than written out
+ * again here, so the scaffold cannot drift from what the catalogue calls
+ * current. Anthropic publishes no embeddings API, and a `--model` naming a
+ * provider we do not ship gives nothing to guess from; both scaffold without
+ * one, and memory still works by ranking recall on term overlap. An
+ * `embeddings:` entry pointing at a provider that has none is not a degraded
+ * project, it is one that fails to load.
+ */
+function defaultEmbedding(provider) {
+    if (provider === undefined) {
+        return undefined;
+    }
+    const id = PREFERRED[provider].embedding[0];
+    return id === undefined ? undefined : `${provider}:${id}`;
+}
+/**
+ * The provider a model ref names, when it names one we know.
+ *
+ * Every ref the scaffold writes carries its provider, so this is how the
+ * memory store ends up on the same credential as the chat model — including
+ * for an explicit `--model`, where nothing else says which vendor was meant.
+ */
+function providerOf(ref) {
+    const colon = ref.indexOf(':');
+    const head = colon < 0 ? '' : ref.slice(0, colon);
+    return isProvider(head) ? head : undefined;
+}
 /**
  * The provider this machine can actually reach.
  *
@@ -96,10 +127,11 @@ export const init = {
     summary: 'Create a project here, or in <dir>, and register it.',
     usage: USAGE,
     details: [
-        'Writes SPECIFICATION.md, INSTRUCTIONS.md, agents.yaml, agents/ and',
-        'scripts/, then records the directory so `zen list` and `zen open` can',
-        'find it by name. Editor files (.vscode/settings.json and the .github/',
-        'tree) are written alongside, replacing any already there.',
+        'Writes SPECIFICATION.md, agents.yaml, agents/ (house rules, the memory',
+        'rules and one role prompt) and scripts/, then records the directory so',
+        '`zen list` and `zen open` can find it by name. Editor files',
+        '(.vscode/settings.json and the .github/ tree) are written alongside,',
+        'replacing any already there.',
         '',
         'A directory that is already a project is adopted rather than rebuilt:',
         'it is registered under its own name, and nothing in it is touched.',
@@ -108,10 +140,12 @@ export const init = {
         '',
         'SPECIFICATION.md states what the project is for and what it may do,',
         'and describes the one that was just written. Say what you are building',
-        'there, then run /sync-with-spec in your editor to make the rest match.',
+        'there, then run /spec-sync-project in your editor to make the rest match.',
         '',
         'The default agent gets the file tools and a sandboxed shell, plus',
-        '`exa:*` when the keyring holds an Exa key. Without --model, the',
+        '`exa:*` when the keyring holds an Exa key. It keeps a memory of its',
+        'own work in memory/, vectorised by the same provider as the model',
+        'when that provider publishes embeddings. Without --model, the',
         'keyring is checked and the model is picked from a credential the',
         'provider accepts.',
         '',
@@ -143,7 +177,8 @@ export const init = {
             : DEFAULT_MODEL[provider ?? 'openai'];
         const model = choice.ref;
         const web = hasExa(store);
-        const written = scaffold({ dir, model, modelOptions: choice.options, web });
+        const embedding = defaultEmbedding(providerOf(model));
+        const written = scaffold({ dir, model, modelOptions: choice.options, web, embedding });
         const registry = await Registry.open();
         registry.add(name, dir);
         registry.save();