atris 3.74.4 → 3.74.5

package/atris/team/opus-overnight/MISSION.md

@@ -1,10 +1,10 @@
-# Mission — opus-overnight
+# Mission: opus-overnight
 
 <!-- Human-authored purpose file. Keep this durable; runtime state belongs in .atris/state/*.jsonl and now.md. -->
 
 ## North Star
 
-**AGI is the mission.** Concretely, the path is: the best model that works with Atris first (Atris-shaped, customer-trajectory-driven, vertically RL'd), then we swap the base for a better one when the loop justifies it. The loop, the data, and the published number are the moat — not the base model.
+**AGI is the mission.** Concretely, the path is: the best model that works with Atris first (Atris-shaped, customer-trajectory-driven, vertically RL'd), then we swap the base for a better one when the loop justifies it. The loop, the data, and the published number are the moat, not the base model.
 
 The motto: **every tick, name the 1%.** Each tick must answer "what's 1% better than last tick, and on what axis?" If a tick can't name its 1%, log it as a no-lift tick honestly and pick a different artifact.
 
@@ -16,25 +16,25 @@ The motto: **every tick, name the 1%.** Each tick must answer "what's 1% better
 
 ALL of:
 1. One concrete artifact landed in `rl-exp2/` (script, doc, eval, dataset, comparison) on a tick-prefixed commit.
-2. **The tick summary names its 1% — what got better, by how much, vs what baseline.** Examples: "+1 training mode unlocked (DPO)", "−1 ambiguity in publishable-number rubric", "+5pp clarity on which base model swap is right", "no-lift tick: tried X, didn't move forward, here's what we learned."
+2. **The tick summary names its 1%: what got better, by how much, vs what baseline.** Examples: "+1 training mode unlocked (DPO)", "−1 ambiguity in publishable-number rubric", "+5pp clarity on which base model swap is right", "no-lift tick: tried X, didn't move forward, here's what we learned."
 3. `npm test` in atris-cli still green if any atris-cli code touched (usually none).
 4. A mission-tick receipt recorded via `atris mission tick <id> --summary "Tick N: ... · 1%: <delta>"` so the operator sees the delta in `atris mission status` without reading my chat.
-5. **Zero dollars spent during the tick** — no firectl create/fire/scale, no GPU rentals, no atrisos-backend master commits.
+5. **Zero dollars spent during the tick**, no firectl create/fire/scale, no GPU rentals, no atrisos-backend master commits.
 
 ## Bounded Scope
 
-- Code/docs in `rl-exp2/` — write freely.
-- Reads from `atrisos-backend/backend/rl/` and `atris/features/rl-gym/` — write findings into `rl-exp2/notes/`, never modify the backend without explicit approval.
-- Supabase reads via the existing trajectory_service shape — never insert/update.
-- `firectl list / get / whoami / show` — read-only fine; `create / scale / delete` forbidden.
+- Code/docs in `rl-exp2/`: write freely.
+- Reads from `atrisos-backend/backend/rl/` and `atris/features/rl-gym/`: write findings into `rl-exp2/notes/`, never modify the backend without explicit approval.
+- Supabase reads via the existing trajectory_service shape: never insert/update.
+- `firectl list / get / whoami / show`: read-only fine; `create / scale / delete` forbidden.
 
 ## Domains In Order
 
-1. **Evaluator surface** — heuristic v1 shipped (commit `ae14e32`); LLM-judge v2 next.
-2. **Corpus** — build_sft_corpus.py shipped (commit `bd171fb`); add eval-prompts companion + a dry-run sample dataset for review.
-3. **Base-model swap analysis** — Kimi K2 vs Qwen3.5-30B vs GLM-4.5-Air vs GPT-OSS-20B on Fireworks: cost, throughput, RFT shape support.
-4. **Strategy doc** — smallest publishable number that lets Atris claim "top RL co for vertical workspaces."
-5. **rl-gym artifact survey** — 192 commits, 166 proof artifacts already in atrisos-backend. Map them so future ticks reuse instead of rebuild.
+1. **Evaluator surface**: heuristic v1 shipped (commit `ae14e32`); LLM-judge v2 next.
+2. **Corpus**: build_sft_corpus.py shipped (commit `bd171fb`); add eval-prompts companion + a dry-run sample dataset for review.
+3. **Base-model swap analysis**: Kimi K2 vs Qwen3.5-30B vs GLM-4.5-Air vs GPT-OSS-20B on Fireworks: cost, throughput, RFT shape support.
+4. **Strategy doc**: smallest publishable number that lets Atris claim "top RL co for vertical workspaces."
+5. **rl-gym artifact survey**: 192 commits, 166 proof artifacts already in atrisos-backend. Map them so future ticks reuse instead of rebuild.
 
 ## Stop Condition
 
@@ -51,11 +51,11 @@ ALL of:
 
 ## Search Discipline (anti-laziness rule)
 
-When the operator says "X exists in our code" or "we have X" and a grep returns 0 hits — **do NOT conclude X doesn't exist.** That conclusion is a recurring failure mode (see `atris/lessons.md` slug `narrow-grep-hides-matches`).
+When the operator says "X exists in our code" or "we have X" and a grep returns 0 hits, **do NOT conclude X doesn't exist.** That conclusion is a recurring failure mode (see `atris/lessons.md` slug `narrow-grep-hides-matches`).
 
 The required search ladder:
-1. `grep -rn "<term>" <repo>/` — whole repo, no subdir gating
+1. `grep -rn "<term>" <repo>/`: whole repo, no subdir gating
 2. If still nothing, try variants: case-insensitive (`-i`), alternative spellings, related terms
-3. Only after BOTH (1) and (2) miss, surface "I can't find it — point me at where" rather than declaring absence
+3. Only after BOTH (1) and (2) miss, surface "I can't find it, point me at where" rather than declaring absence
 
 Operator's trust is more reliable than my regex. Trust them, then go find it.

package/atris/team/opus-overnight/STEERING.md

@@ -1,4 +1,4 @@
-# STEERING — opus-overnight
+# STEERING: opus-overnight
 
 <!-- Operator drops one-line directives here while the autonomous loop runs.
      Each tick reads this file FIRST, applies any unapplied directive as
@@ -8,7 +8,7 @@
 
 <!-- Drop one-line directives below. Format: `- YYYY-MM-DD: <directive>`.
      Examples and usage live in rl-exp2/notes/operator-steering.md.
-     Do NOT put examples here — every `- ` line under `## Pending` is read
+     Do NOT put examples here: every `- ` line under `## Pending` is read
      as a real directive on the next tick. -->
 
 ## Applied
@@ -28,8 +28,8 @@ The autonomous loop runs without the operator present. They can't redirect mid-t
 
 This is the lightest layer of the four-tier feedback design:
 - **Layer 1 (this file):** taste tweaks, one-line drops
-- **Layer 2:** `atris brain edit <tick> "..."` — thumbs-up/down on past ticks
-- **Layer 3:** `atris mission steer <id> --note "..."` — directives attached to mission state (not yet implemented)
-- **Layer 4:** edit MISSION.md / MEMBER.md — durable contract changes
+- **Layer 2:** `atris brain edit <tick> "..."`: thumbs-up/down on past ticks
+- **Layer 3:** `atris mission steer <id> --note "..."`: directives attached to mission state (not yet implemented)
+- **Layer 4:** edit MISSION.md / MEMBER.md: durable contract changes
 
 Use the lightest tier that fits. Most steering belongs in Layer 1.

package/atris.md

@@ -55,6 +55,15 @@ In a project with nothing on the list yet, speak first: three things you noticed
 
 You can move fast. You do not get to move blindly.
 
+We have done a lot of great work, and because of it we will do even more.
+
+- **Acceleration.** Progress compounds. The early days are slow, then the curve bends and everything speeds up. Build on what is already done and settled; never redo it or re-ask it. Each landed piece makes the next one faster.
+- **Accuracy.** A win counts when the result is validated: a test, a gate, or a real run proves it. Small fast tasks can be proven by a check alone; good and big work also gets a second reviewer. Accuracy is the floor the other three stand on, so better tests make everything faster.
+- **Ambition.** Always one more level. When the goal has a clear check, go for the bigger version or the great task. We are capable of it, and we always find ways to win.
+- **Agency.** Take action and own the outcome. Use every tool you have. When something is missing or broken, fix it the same turn instead of handing it back.
+
+Acknowledge each validated win, then go for the next one.
+
 Scale ceremony to blast radius. Before a nontrivial change (new surface,
 multi-file, migrations, anything hard to roll back), state:
 - the goal

package/bin/atris.js

@@ -143,8 +143,8 @@ const helpRequested = updateCommand === 'help'
 const jsonRequested = process.argv.slice(2).includes('--json');
 const dryRunRequested = updateArgs.includes('--dry-run');
 const skipUpdateCheck = Boolean(process.env.ATRIS_SKIP_UPDATE_CHECK || process.env.NO_UPDATE_NOTIFIER || helpRequested || jsonRequested);
-// recap promises local only: no registry call and no background install.
-if (!skipUpdateCheck && (!updateCommand || (updateCommand && !['version', 'update', 'mcp', 'recap'].includes(updateCommand)))) {
+// Local-only commands make no registry call and start no background install.
+if (!skipUpdateCheck && (!updateCommand || (updateCommand && !['version', 'update', 'mcp', 'recap', '99'].includes(updateCommand)))) {
   updateCheckPromise = checkForUpdates()
     .then((updateInfo) => {
       if (updateInfo) {
@@ -528,6 +528,7 @@ function showHelpShort() {
   console.log('');
   console.log('  atris "what do you want here?"   empty folder');
   console.log('  atris later "..."                remember without a room');
+  console.log('  atris idea "..."                 drop an idea; atris builds it later');
   console.log('  atris guide "..."                translate your words');
   console.log('  atris do                         work already here');
   console.log('  atris done                       land what your agent built');
@@ -563,6 +564,7 @@ function showHelpAll() {
   console.log('you say what you want. already won. one next step.');
   console.log('');
   console.log('  atris later "..."                remember without a room');
+  console.log('  atris idea "..."                 drop an idea; atris builds it later');
   console.log('  atris guide "..."                translate your words');
   console.log('  atris do                         work already here');
   console.log('  atris done                       land what your agent built');
@@ -600,6 +602,7 @@ function showHelpAll() {
   console.log('  log        - Add ideas to inbox (`atris log "note"`; `log --repl` for the editor; see also `wish`)');
   console.log('  wish       - Say one plain sentence, then Atris asks only for gaps or delegates it');
   console.log('  now        - Show atris/now.md, the current operating truth');
+  console.log('  99         - focus, meditation, health, and optional sync');
   console.log('  goal       - goal, distance, and what moved today (alias: wtf)');
   console.log('  founder    - show weekly commit and completed-task improvement from receipts');
   console.log('  orb        - Pick next moves while engine jobs work in the background');
@@ -608,7 +611,7 @@ function showHelpAll() {
   console.log('  who        - Show local engines and team members working, waiting, done, or stale');
   console.log('  stream     - Watch the whole team work live in one terminal');
   console.log('  team       - One team view: members, roles, engine assignments (presence for live)');
-  console.log('  gm         - Play your team as a front office: ratings, stuck jobs, league, replay, phone');
+  console.log('  gm         - Play your team as a front office: ratings, stuck jobs, league, replay, phone, walk');
   console.log('  host       - The Host: gets to know people, makes two-yes introductions (atris member install host)');
   console.log('  watch      - Turn one sentence into an always-on background watcher');
   console.log('  ctop       - Show a process-first live agent CPU/memory view');
@@ -2288,6 +2291,9 @@ if (command === 'guide') {
     console.error(error.message || String(error));
     process.exit(1);
   }
+} else if (command === '99') {
+  Promise.resolve(require('../commands/ninety-nine').ninetyNineAtris(process.argv.slice(3)))
+    .then((code) => exitAfterStdoutDrain(code));
 } else if (command === 'now') {
   const args = process.argv.slice(3);
   const nowCode = require('../commands/now').nowAtris(args);
@@ -2298,6 +2304,13 @@ if (command === 'guide') {
 } else if (command === 'later') {
   const laterCode = require('../commands/later').laterAtris(process.argv.slice(3));
   process.exit(Number.isInteger(laterCode) ? laterCode : 0);
+} else if (command === 'idea') {
+  Promise.resolve(require('../commands/idea').ideaAtris(process.argv.slice(3))).then((ideaCode) => {
+    process.exit(Number.isInteger(ideaCode) ? ideaCode : 0);
+  }, (error) => {
+    console.error(`atris idea: ${error && error.message || error}`);
+    process.exit(1);
+  });
 } else if (command === 'done') {
   const doneCode = require('../commands/done').doneAtris(process.argv.slice(3));
   process.exit(Number.isInteger(doneCode) ? doneCode : 0);

package/commands/campaign.js

@@ -33,7 +33,7 @@ function printHelp() {
   console.log('  atris campaign start "<title>" --judge "<cmd>" [--budget <dollars>]   no --budget means no limit; --cloud needs one');
   console.log('        [--holdout <file>]   move held-out data out of the repo; the judge reads it from $ATRIS_CAMPAIGN_HOLDOUT');
   console.log('        [--owner <member>] [--seat <member>[:<role>]]... [--bet "<bet>" [--wrong-if "<what proves it wrong>"]]');
-  console.log('        [--standing] [--parent <id>] [--id cmp-<name>] [--cloud [--basis atris_cloud]] [--protect <path>]... [--run]');
+  console.log('        [--standing] [--parent <id>] [--id cmp-<name>] [--cloud [--basis atris_cloud]] [--protect <path>]... [--brief <file>] [--run]');
   console.log('        --protect saves paths agents must not touch, and the commit they start from');
   console.log('        --basis atris_cloud: Atris runs it on your Atris computer and bills its engines; the default is your own engines');
   console.log('  atris campaign show <id>               bet, judge, answer, budget, seats, missions, asks');
@@ -89,7 +89,7 @@ const NO_LIMIT_BUDGET = 1e9;
 const noLimit = (view) => view.budget.total >= NO_LIMIT_BUDGET;
 
 // The flags this command understands, so positional words are what is left.
-const VALUE_FLAGS = ['--mission-minutes', '--basis', '--work', '--break', '--rounds', '--mission', '--parallel', '--max-steps', '--max-ticks', '--runner', '--count', '--judge', '--budget', '--owner', '--seat', '--bet', '--wrong-if', '--parent', '--id', '--role', '--add', '--data', '--timeout', '--as', '--holdout', '--protect', '--every', '--steps'];
+const VALUE_FLAGS = ['--mission-minutes', '--basis', '--work', '--break', '--rounds', '--mission', '--parallel', '--max-steps', '--max-ticks', '--runner', '--count', '--judge', '--budget', '--owner', '--seat', '--bet', '--wrong-if', '--parent', '--id', '--role', '--add', '--data', '--timeout', '--as', '--holdout', '--protect', '--brief', '--every', '--steps'];
 
 function positionals(args) {
   const out = [];
@@ -389,6 +389,14 @@ async function cmdStart(args, ctx) {
   const protectPlan = protect.length ? planProtect(process.cwd(), protect) : null;
   if (protectPlan && protectPlan.error) return ctx.fail(protectPlan.error);
 
+  const briefFlag = readFlag(args, '--brief', '');
+  let briefText = null;
+  if (briefFlag) {
+    const from = path.resolve(briefFlag);
+    if (!fs.existsSync(from) || !fs.statSync(from).isFile()) return ctx.fail(`--brief ${briefFlag} is not a file`);
+    briefText = fs.readFileSync(from, 'utf8');
+  }
+
   const holdoutFlag = readFlag(args, '--holdout', '');
   let held = null;
   if (holdoutFlag) {
@@ -425,6 +433,12 @@ async function cmdStart(args, ctx) {
   if (parent) ctx.append(id, 'attached', { parent_id: parent });
   ctx.append(id, 'judge_ran', judgeRanData(reading), 'judge');
   if (protectPlan) saveProtect(ctx.campaignHome, id, protectPlan);
+  if (briefText !== null) {
+    const { briefPath } = require('../lib/campaign-loop');
+    const target = briefPath(process.cwd(), id);
+    fs.mkdirSync(path.dirname(target), { recursive: true });
+    fs.writeFileSync(target, briefText);
+  }
 
   let cloud = null;
   if (args.includes('--cloud')) {
@@ -800,6 +814,7 @@ async function driveLoop(args, ctx, verb, view) {
     log: ctx.asJson ? () => {} : (line) => console.log(line),
     judge: (command) => ctx.judge(command, JUDGE_TIMEOUT_SECONDS, view.id),
     lessons: (id) => require('../lib/campaign-cadence').readLessons(process.cwd(), id),
+    brief: (id) => require('../lib/campaign-loop').readBrief(process.cwd(), id),
     ...(sync ? { sync } : {}),
     ...(ctx.loopDeps || {}),
   });

package/commands/card.js

@@ -11,6 +11,7 @@ const path = require('path');
 const os = require('os');
 const { execFileSync, spawnSync } = require('child_process');
 const { buildCard, SIZES, KINDS } = require('../lib/card');
+const { makeProfileDir, removeProfileDir, headlessArgs } = require('../lib/headless-chrome');
 
 function parseFlags(argv) {
   const flags = {}; const pos = [];
@@ -46,18 +47,24 @@ function findChrome() {
 
 function renderPng(chrome, html, width, height, outPng) {
   const tmp = fs.mkdtempSync(path.join(os.tmpdir(), 'atris-card-'));
-  const htmlFile = path.join(tmp, 'card.html');
-  fs.writeFileSync(htmlFile, html);
-  const args = [
-    '--headless=new', '--disable-gpu', '--hide-scrollbars',
-    '--force-device-scale-factor=2',
-    `--window-size=${width},${height}`,
-    '--virtual-time-budget=4000',
-    `--screenshot=${outPng}`,
-    `file://${htmlFile}`,
-  ];
-  execFileSync(chrome, args, { stdio: 'ignore', timeout: 60000 });
-  return fs.existsSync(outPng);
+  const profile = makeProfileDir('atris-card-chrome-');
+  try {
+    const htmlFile = path.join(tmp, 'card.html');
+    fs.writeFileSync(htmlFile, html);
+    const args = headlessArgs(profile, [
+      '--disable-gpu', '--hide-scrollbars',
+      '--force-device-scale-factor=2',
+      `--window-size=${width},${height}`,
+      '--virtual-time-budget=4000',
+      `--screenshot=${outPng}`,
+      `file://${htmlFile}`,
+    ]);
+    execFileSync(chrome, args, { stdio: 'ignore', timeout: 60000 });
+    return fs.existsSync(outPng);
+  } finally {
+    removeProfileDir(profile);
+    removeProfileDir(tmp);
+  }
 }
 
 function run(argv) {

package/commands/engine.js

@@ -11,6 +11,8 @@
 //   atris engine roster confirm  renew every dated worker for 30 days
 //   atris engine roster session [clear]  this shell's own changes
 //   atris engine roster --available  tools and models on this machine
+//   atris engine roster set <job> <tool> <model>  swap a job's lead; the old lead backs it up
+//   atris engine roster set --member <name> <tool> <model>  give one member its own model
 //   atris engine assign <job> <tool> [--like <kind>] [--model <m>] [--effort <level>] [--max "20 min"] [--prep <job>] [--backup "<tool> [model]"] [--days <n>] [--add] [--session | --everywhere]
 //   atris engine assign <job> --promote "<tool> [model]" | --remove <tool> | --clear [--session | --everywhere]
 //   jobs are open-ended: search, build, review, or your own ("small build")
@@ -35,11 +37,12 @@ const {
 const { parseScopeFlag } = require('../lib/cli-scope');
 const { isFreshWorkspace, speakFirstMinute } = require('../lib/first-minute');
 const { teamRosterView } = require('../lib/member-engine');
-const { engineModelText, engineRunsView, availableModels, modelLabel } = require('../lib/roster-models');
+const { engineModelText, engineRunsView, modelLabel } = require('../lib/roster-models');
 const { renderTable } = require('../lib/text-table');
 const { TEAM_FOOTER } = require('../lib/team-lineup');
 const { readRosterRuns, workerRuns, summarizeRuns, recentRuns, renderRunLine, runView, DEFAULT_DAYS: RUN_DAYS } = require('../lib/roster-runs');
 const { attachSuggestions } = require('../lib/roster-suggest');
+const { swapList, swapJob, swapMember } = require('../lib/roster-swap');
 const { PIN_MAX_AGE_DAYS: PIN_STALE_DAYS, attachStaleModels, modelCatalog } = require('../lib/roster-stale');
 const {
   ENGINE_ROLES,
@@ -48,7 +51,7 @@ const {
   ROSTER_LAYERS,
   rosterDecided,
   rosterWorkerWalk,
-  rosterToolLabel,
+  rosterToolName,
   rosterSessionKey,
   clearSessionRoster,
   NO_SESSION_MESSAGE,
@@ -1399,18 +1402,17 @@ function printJobRoster(root, now = new Date()) {
   console.log(renderRosterReport(rosterReport(root, now)));
 }
 
-// Each installed tool and the models it offers, from local files only.
-function availableReport(root) {
-  const engines = readEngineRegistry(root, { persist: false }).engines
-    .filter((engine) => !HIDDEN_PROFILE_NAMES.includes(engine.id))
-    .filter((engine) => engine.installed || (engine.health && engine.health.status === 'ready'));
-  return availableModels(engines).map((row) => ({ ...row, tool: rosterToolLabel(row.engine) }));
-}
-
-function renderAvailable(rows) {
+// With the owner's picks: the swap list on top (atris engines, then the
+// picks), then every model under "more models". Without: every tool's list.
+function renderAvailable(rows, list = null) {
   if (!rows.length) return 'no tools found on this machine. run atris engine doctor to check again';
-  const width = Math.max(...rows.map((row) => row.tool.length));
-  return ['on this machine', ...rows.map((row) => `${row.tool.padEnd(width)}  ${row.models.join(', ') || 'its own default'}  (${row.source})`)].join('\n');
+  const top = list && list.picked ? list.options.filter((option) => !option.more) : [];
+  const width = Math.max(...rows.map((row) => row.tool.length), ...top.map((option) => option.tool.length));
+  const full = rows.map((row) => `${row.tool.padEnd(width)}  ${row.models.join(', ') || 'its own default'}  (${row.source})`);
+  if (!top.length) return ['on this machine', ...full].join('\n');
+  const labels = Math.max(...top.map((option) => option.label.length));
+  const picks = top.map((option) => `${option.tool.padEnd(width)}  ${option.label.padEnd(labels)}  ${option.atris ? 'atris' : 'pick'}`);
+  return ['on this machine', ...picks, '', 'more models', ...full].join('\n');
 }
 
 // Only this session's changes, from its own file.
@@ -1518,7 +1520,78 @@ function runRosterRunsCommand(job, json, root, now) {
   return 0;
 }
 
+const ROSTER_SET_USAGE = 'usage: atris engine roster set <job> <tool> <model>, or atris engine roster set --member <name> <tool> <model>. the tools and models come from atris engine roster --available';
+
+// The words after set: the job (one quoted argument, or every word before
+// the first one that names a tool), the tool (one word, or two like "claude
+// code"), then the model, which may be several words.
+function splitSetWords(words, needJob) {
+  // "claude code" and "atris fast" are two words; every other tool is one.
+  const toolAt = (i) => {
+    const two = i + 1 < words.length ? `${words[i]} ${words[i + 1]}` : '';
+    if (two && rosterToolNameOf(two)) return { at: i, size: 2 };
+    return rosterToolNameOf(words[i]) ? { at: i, size: 1 } : null;
+  };
+  let start = 0;
+  let job = '';
+  if (needJob) {
+    if (!words.length) return null;
+    if (/\s/.test(String(words[0]).trim())) {
+      job = String(words[0]).trim();
+      start = 1;
+    } else {
+      let found = null;
+      for (let i = 1; i < words.length && !found; i += 1) found = toolAt(i);
+      if (!found) return null;
+      job = words.slice(0, found.at).join(' ');
+      start = found.at;
+    }
+  }
+  const tool = toolAt(start);
+  if (!tool) return words.length > start ? { job, tool: words[start], model: words.slice(start + 1).join(' ') } : null;
+  return { job, tool: words.slice(start, start + tool.size).join(' '), model: words.slice(start + tool.size).join(' ') };
+}
+
+function rosterToolNameOf(words) {
+  return rosterToolName(String(words || '').trim());
+}
+
+function runRosterSetCommand(args, root) {
+  const json = args.includes('--json');
+  const words = args.filter((arg) => arg !== '--json');
+  let member = '';
+  const at = words.indexOf('--member');
+  if (at !== -1) {
+    member = words[at + 1] || '';
+    words.splice(at, 2);
+    if (!member || member.startsWith('--')) {
+      console.error(ROSTER_SET_USAGE);
+      return 2;
+    }
+  }
+  if (words.some((word) => String(word).startsWith('--'))) {
+    console.error(ROSTER_SET_USAGE);
+    return 2;
+  }
+  const parts = splitSetWords(words, !member);
+  if (!parts || (!member && !parts.job)) {
+    console.error(ROSTER_SET_USAGE);
+    return 2;
+  }
+  try {
+    const out = member
+      ? swapMember(root, member, parts.tool, parts.model)
+      : swapJob(root, parts.job, parts.tool, parts.model);
+    console.log(json ? JSON.stringify(out, null, 2) : out.line);
+    return 0;
+  } catch (err) {
+    console.error(err.message);
+    return 2;
+  }
+}
+
 function runRosterCommand(args, root, now = new Date()) {
+  if (args[0] === 'set') return runRosterSetCommand(args.slice(1), root);
   const json = args.includes('--json');
   const available = args.includes('--available');
   const runsAt = args.indexOf('--runs');
@@ -1536,12 +1609,14 @@ function runRosterCommand(args, root, now = new Date()) {
     || (rest.length === 1 && ['confirm', 'session'].includes(rest[0]))
     || (rest.length === 2 && rest[0] === 'session' && rest[1] === 'clear');
   if (!known || (available && rest.length)) {
-    console.error('usage: atris engine roster [confirm | session [clear] | --available | --runs [job]] [--json]');
+    console.error('usage: atris engine roster [confirm | session [clear] | set <job> <tool> <model> | set --member <name> <tool> <model> | --available | --runs [job]] [--json]');
     return 2;
   }
   if (available) {
-    const rows = availableReport(root);
-    console.log(json ? JSON.stringify({ tools: rows }, null, 2) : renderAvailable(rows));
+    // the same ordered list the phone's engine card reads
+    const list = swapList(root);
+    const rows = list.tools;
+    console.log(json ? JSON.stringify({ tools: rows, list: list.options }, null, 2) : renderAvailable(rows, list));
     return 0;
   }
   if (rest[0] === 'session') return runRosterSessionCommand(rest, json, root, now);
@@ -2548,7 +2623,7 @@ function engineCommand(args = [], deps = {}) {
     if (sub === 'help' || args.includes('--help') || args.includes('-h')) {
       console.log('\n  atris engine watch [<id>|latest] [--no-follow]\n                           follow one live transcript or list running engine work');
       console.log('\n  long read-only asks: atris engine ask "..." --engine agy --timeout <seconds>\n                           quick asks default to 120 seconds; explicit jobs allow up to 3600\n                           follow live work with atris engine watch latest');
-      console.log('\n  atris engine            roster + current default\n  atris engine roster [--json] show every job and its workers in order: search, build, review, and your own\n                           flags a model with a newer version on this machine, or a pin not proven in 30 days\n                           (a landed run, or "verified 2026-09-30" on its line, proves it)\n  atris engine roster confirm renew every dated worker for 30 days\n  atris engine roster session [clear]\n                           show or drop the roster changes for this shell only\n  atris engine roster --available\n                           tools on this machine and the models each one offers\n  atris engine roster --runs [job]\n                           the last 20 runs, newest first, with tool, model, time, and outcome\n  atris engine assign <job> <tool> [--like <kind>] [--model <m>] [--effort <level>] [--max "20 min"] [--prep <job>] [--backup "<tool> [model]"] [--days <n>] [--add] [--session | --everywhere]\n  atris engine assign <job> --promote "<tool> [model]" | --remove <tool> | --clear [--session | --everywhere]\n                           sets the lead worker of the job; --promote moves one of its workers to the top as written;\n                           --add puts one more at the end; jobs: search, build, review,\n                           or your own like "small build" (--like search|build|review when the name does not say);\n                           --prep search has that job read the task first and hand this worker a short brief;\n                           --everywhere is for all projects; --session is this shell only (set ATRIS_ROSTER_SESSION=<name>)\n  atris engines --chart   show the fleet as an org chart\n  atris engine list --json [--all] full registry: default + engines with tier, roles, fallback, health (--all includes hidden engines)\n  atris engine set <name> --duty leader|errands|learning [--models "a, b"]\n                           arrange the fleet and save its model policy\n  atris engine resolve <role> or <job> [--json]\n                           choose the best ready engine for navigator|executor|validator or a roster job;\n                           notes always answers: its roster pick, else agy then haiku\n  atris engine health <name> --set ready|not_installed|credit_out\n                           flip runtime health, for example when credits run out\n  atris engine doctor [--json]\n                           probe which engine CLIs are installed here and sync that into health policy\n  atris engine <name>     make that engine the default here\n  atris engine test [name] preflight: run the engine CLI headless, report pass/fail\n  atris engine bench [names...] [--runs N]\n                           ranked latency scoreboard of engine passes\n  atris engine ask "<question>" --engine <name> [--engine <name> ...]\n                           ask several engines in parallel without allowing edits\n  atris engine ask --jobs <jobs.json>\n                           ask different read-only questions in parallel\n  atris engine validate <receipt-path|latest> [--engine <name>]\n                           check ask answers with a different read-only referee\n  atris engine validate scoreboard\n                           show pass rates by worker engine\n  atris engine sweep [--apply] [--json]\n                           list ssh-agents that engine shells left running; --apply rechecks and stops them\n  atris engine dispatch <task-id> [<task-id> ...] --engine cursor|codex [--prompt-file <f>] [--yolo]\n                           one-command claim, worktree, build, verify, ship, ready\n  atris engine login <provider> --yes\n                           upload a local provider CLI login to the backend vault\n  atris engine login <provider> --computer [--seat <name>]\n  atris engine login <provider> --business <id> [--seat <name>]\n                           sign in on an Atris computer by device flow\n  atris engine login --list | --remove <provider>\n                           list or remove vaulted provider logins\n  atris engine seats       show which named accounts are ready to work\n  atris engine seed <provider> --business <id>|--user\n                           push a vaulted login onto an Atris computer\n  atris engine reset      back to the house default\n  --engine <name>         one run on that engine (mission run / autopilot / run)\n');
+      console.log('\n  atris engine            roster + current default\n  atris engine roster [--json] show every job and its workers in order: search, build, review, and your own\n                           flags a model with a newer version on this machine, or a pin not proven in 30 days\n                           (a landed run, or "verified 2026-09-30" on its line, proves it)\n  atris engine roster confirm renew every dated worker for 30 days\n  atris engine roster session [clear]\n                           show or drop the roster changes for this shell only\n  atris engine roster --available\n                           tools on this machine and the models each one offers\n  atris engine roster set <job> <tool> <model>\n                           swap the lead of a job to a tool and model this machine offers; the old lead becomes the first backup\n  atris engine roster set --member <name> <tool> <model>\n                           give one member its own model; every other job stays put\n  atris engine roster --runs [job]\n                           the last 20 runs, newest first, with tool, model, time, and outcome\n  atris engine assign <job> <tool> [--like <kind>] [--model <m>] [--effort <level>] [--max "20 min"] [--prep <job>] [--backup "<tool> [model]"] [--days <n>] [--add] [--session | --everywhere]\n  atris engine assign <job> --promote "<tool> [model]" | --remove <tool> | --clear [--session | --everywhere]\n                           sets the lead worker of the job; --promote moves one of its workers to the top as written;\n                           --add puts one more at the end; jobs: search, build, review,\n                           or your own like "small build" (--like search|build|review when the name does not say);\n                           --prep search has that job read the task first and hand this worker a short brief;\n                           --everywhere is for all projects; --session is this shell only (set ATRIS_ROSTER_SESSION=<name>)\n  atris engines --chart   show the fleet as an org chart\n  atris engine list --json [--all] full registry: default + engines with tier, roles, fallback, health (--all includes hidden engines)\n  atris engine set <name> --duty leader|errands|learning [--models "a, b"]\n                           arrange the fleet and save its model policy\n  atris engine resolve <role> or <job> [--json]\n                           choose the best ready engine for navigator|executor|validator or a roster job;\n                           notes always answers: its roster pick, else agy then haiku\n  atris engine health <name> --set ready|not_installed|credit_out\n                           flip runtime health, for example when credits run out\n  atris engine doctor [--json]\n                           probe which engine CLIs are installed here and sync that into health policy\n  atris engine <name>     make that engine the default here\n  atris engine test [name] preflight: run the engine CLI headless, report pass/fail\n  atris engine bench [names...] [--runs N]\n                           ranked latency scoreboard of engine passes\n  atris engine ask "<question>" --engine <name> [--engine <name> ...]\n                           ask several engines in parallel without allowing edits\n  atris engine ask --jobs <jobs.json>\n                           ask different read-only questions in parallel\n  atris engine validate <receipt-path|latest> [--engine <name>]\n                           check ask answers with a different read-only referee\n  atris engine validate scoreboard\n                           show pass rates by worker engine\n  atris engine sweep [--apply] [--json]\n                           list ssh-agents that engine shells left running; --apply rechecks and stops them\n  atris engine dispatch <task-id> [<task-id> ...] --engine cursor|codex [--prompt-file <f>] [--yolo]\n                           one-command claim, worktree, build, verify, ship, ready\n  atris engine login <provider> --yes\n                           upload a local provider CLI login to the backend vault\n  atris engine login <provider> --computer [--seat <name>]\n  atris engine login <provider> --business <id> [--seat <name>]\n                           sign in on an Atris computer by device flow\n  atris engine login --list | --remove <provider>\n                           list or remove vaulted provider logins\n  atris engine seats       show which named accounts are ready to work\n  atris engine seed <provider> --business <id>|--user\n                           push a vaulted login onto an Atris computer\n  atris engine reset      back to the house default\n  --engine <name>         one run on that engine (mission run / autopilot / run)\n');
       return 0;
     }
     if (isFreshWorkspace(root)) {

package/commands/gm.js

@@ -31,6 +31,7 @@ function showHelp() {
   console.log('       atris gm card <member> [--json]  one member\'s profile page');
   console.log('       atris gm target [--landed N] [--max-abandoned N] [--max-stuck N] [--clear]');
   console.log('       atris gm phone [--port 7777] [--send] [--new-link]  decision cards on your phone');
+  console.log('       atris gm walk [--real-copy <workspace>] [--threshold 0.8] [--no-browser]  check every stage of the phone orb');
   console.log('       atris gm --json           machine-readable state');
   console.log('       atris gm --watch [--interval 10]');
   console.log('');
@@ -121,6 +122,13 @@ function showHelp() {
   console.log('  tap. Every tap is logged to ~/.atris/state/gm-phone-actions.jsonl.');
   console.log('  --send texts the link to the number saved with atris autoland on --to.');
   console.log('');
+  console.log('  atris gm walk starts the phone page on a made-up workspace with a temp');
+  console.log('  home, walks every stage of the orb (the idea, an answer, the weekly');
+  console.log('  number, the team, a swap, the job deck, a yes card, Esc, the small and');
+  console.log('  big orb), and prints one row per stage: pass or fail, and why. Rules in');
+  console.log('  code and a fast judge model both have to agree. atris gm walk --help');
+  console.log('  says more. It runs from an atris checkout and never touches ~/.atris.');
+  console.log('');
   console.log('  The league treats every real project folder with task events in the');
   console.log('  last 30 days as a club: team rating, landed this week and in 30 days,');
   console.log('  drops, abandons, stuck claims, last task event, and best player.');
@@ -1622,6 +1630,22 @@ async function gmPhone(args = []) {
   return undefined;
 }
 
+// ---- walk ----------------------------------------------------------------------
+
+// gm walk: every stage of the phone orb, walked and judged on a made-up
+// workspace (scripts/orb-walk.js). The published package leaves the script
+// out, so an install says so instead of failing on a missing file.
+async function gmWalk(args = []) {
+  const script = path.join(__dirname, '..', 'scripts', 'orb-walk.js');
+  if (!fs.existsSync(script)) {
+    console.error('atris gm walk runs from an atris checkout. this install does not include the walk.');
+    process.exitCode = 2;
+    return undefined;
+  }
+  process.exitCode = await require(script).main(args);
+  return undefined;
+}
+
 // ---- player card ------------------------------------------------------------
 
 // One member's profile page as lines. Pure: card in, lines out.
@@ -1844,6 +1868,12 @@ function wakeableMember(args) {
 }
 
 async function gmCommand(...args) {
+  // the walk has its own help
+  if (positional(args)[0] === 'walk') {
+    const at = args.indexOf('walk');
+    return gmWalk([...args.slice(0, at), ...args.slice(at + 1)]);
+  }
+
   if (args.includes('--help') || args.includes('-h') || args[0] === 'help') {
     showHelp();
     return;