Skip to content

Open-source platform for 3D AI agents. Turn text or a photo into a rigged, animated GLB avatar, give it an LLM brain, memory and a wallet, and embed it anywhere with one web component. Remote MCP server, x402 payments, Three.js, browser-native.

docs(owner-actions): re-measure the deploy, sponsor and RPC rows from the 2026-10-09 triage sweep and add the GPU worker redeploy row

nirholas52c86bfe8ef6
diff --git a/prompts/finish/_context/production-100-OWNER-ACTIONS.md b/prompts/finish/_context/production-100-OWNER-ACTIONS.md
index 9e387ff81..a6f0eb72a 100644
--- a/prompts/finish/_context/production-100-OWNER-ACTIONS.md
+++ b/prompts/finish/_context/production-100-OWNER-ACTIONS.md
@@ -9,12 +9,12 @@ it and, if that unblocks a work order, say which one in the owning pack's `produ
Rows carry the date they were added so staleness is visible. Re-verify a row before acting
on it; some resolve themselves as other work lands.
-## Open rows (2026-10-09: row 30 added for the Grok Bot CLI setup rerun; 2026-10-08: row 26 added for the coin-gated MCP catalog regeneration; rows 1 and 2 re-measured by the triage sweep, row 24 added for the paid Solana RPC credits, row 25 added for the MCP connector OAuth deploy; 2026-09-30: row 23 added for the $THREE receiver wallets; 2026-09-29: row 19 re-measured after the agent-sniper deploy; 2026-09-09: row 19 added, and row 1 corrected because its outage premise no longer holds; rows 1, 2, 3, 9, 13, 15 and 16 re-measured 2026-09-04; the former row 18 was cleared that day and its number is not reused; the 2026-09-02 and 2026-08-09 passes are folded in below)
+## Open rows (2026-10-09: rows 1, 2 and 24 re-measured by the triage sweep and row 31 added for the GPU worker redeploy; row 30 added for the Grok Bot CLI setup rerun; 2026-10-08: row 26 added for the coin-gated MCP catalog regeneration; rows 1 and 2 re-measured by the triage sweep, row 24 added for the paid Solana RPC credits, row 25 added for the MCP connector OAuth deploy; 2026-09-30: row 23 added for the $THREE receiver wallets; 2026-09-29: row 19 re-measured after the agent-sniper deploy; 2026-09-09: row 19 added, and row 1 corrected because its outage premise no longer holds; rows 1, 2, 3, 9, 13, 15 and 16 re-measured 2026-09-04; the former row 18 was cleared that day and its number is not reused; the 2026-09-02 and 2026-08-09 passes are folded in below)
| # | Action | Size | Unblocks | Detail lives in |
|---|---|---|---|---|
-| 1 | Approve the production deploy (or say "get production working", which is itself the approval). **The outage this row was written for is CLEARED.** Re-measured 2026-09-09 from the live origin: production serves `880bdcef8` on revision `three-ws-api-00420-ljh`, built 2026-09-08 19:05 UTC, and the `ERR_MODULE_NOT_FOUND` on `/app/services/home-relay/src/token.js` is gone: `/api/healthz` answers **HTTP 200** and `/api/home/status` and `/api/home/connections` answer 401 (auth required) rather than 500. What remains is ordinary staleness, 182 commits behind `main`, which is a routine cadence question and no longer a platform-health emergency. The 2026-09-04 text of this row, which described 96 commits and a 500ing healthz, is superseded. **Re-measured 2026-10-08 17:05 UTC:** production serves `fe2a8b24f` on `three-ws-api-00479-d7z`, **146 commits behind `main`, 90 of them fixes** (security, payouts, SSRF and XSS among them). Shipping them clears four live findings at once: the ring sweep's fee falls back to the treasury instead of the dry sponsor (`a9c996e72`, about 30 failed sweeps an hour since 2026-10-03), an EVM testnet identity-index cursor frozen since 2026-09-11 (`21fa42bb5`), two pages and their two docs pages advertised in the sitemap but answering 404 (added in `16f9d07f4`), and `/tty` holding its animation until the end for any client that accepts compression (`03b4f1189`). One migration must go with it: `20261008150000_revenue_direct_settle_backfill.sql` (pending; it matches 0 rows today, so applying it is safe, and the code half is what keeps the payout cron from paying a direct-to-seller sale twice). Commands, in order: `npm run db:status`, `npm run db:migrate`, `npm run deploy:gcp:full`. | one yes | Shipping whatever has landed on `main` since 2026-09-08 | [01-ship-readiness.md](../901-production-100-01-ship-readiness.md) |
-| 2 | Fund the settle sponsor wallet `WwwuGbqHrwF5RG89KhUbmRWEvjnRH9k5kVM5p7T3WwW`. Healthz reports `x402_settle` down with `cause: sponsor_floor` (settle 5.9%, 3 of 51 paid attempts over 3 hours, 329 `no_solana_accept`) and the sniper worker out of capital (0.2307 SOL to refill its wallets, funding master holds 0.0250 SOL); no config change moves either. Re-read 2026-09-04 08:21 UTC against mainnet: the sponsor holds **0.001507661 SOL** (1,507,661 lamports), still below the settle floor, so the Solana accept stays withdrawn from every 402 challenge. Two months of re-measurement have only ever moved this number down; nothing on this machine can move it up. **Re-measured 2026-10-08 17:05 UTC:** the sponsor holds **651,828 lamports (0.000652 SOL)**, below rent-exempt (890,880), and healthz reads `x402_settle` down at 0.0% (0 of 32 paid attempts in 3 hours, 58 settles refused at the fee-wallet floor). Because the sponsor cannot co-sign, every 402 challenge advertises the Solana accept without `feePayer`, and the ring skips it: 151 `missing_fee_payer` skips in 24 hours on the MCP streaming health probe alone. Last successful settle: 2026-10-03 19:24 UTC. **Row 1's deploy alone restores the ring, no funding needed** (diagnosed 2026-10-08 by read-only simulation): the treasury pays its own sweep fee (5,001 lamports; it holds 0.060 SOL, about 11,800 sweeps) and moves its stranded 2.587 USDC back to the ring payer, which holds 341 atomics today, and ring settles are self-pay. What still needs this wallet: the sponsor-mode callers (the loop's inline entries and the MCP streaming health probe) and the facilitator's `sponsor_fee_unfunded` verify rejects. Clearing the 2,000,000-lamport floor takes at least **1,348,172 lamports**. | ~1 SOL | x402 settle recovery, backlog 01 retirement, sniper | [../backlog/PROGRESS.md](backlog-PROGRESS.md), 2026-08-02 and 2026-09-01 entries |
+| 1 | Approve the production deploy (or say "get production working", which is itself the approval). **The outage this row was written for is CLEARED.** Re-measured 2026-09-09 from the live origin: production serves `880bdcef8` on revision `three-ws-api-00420-ljh`, built 2026-09-08 19:05 UTC, and the `ERR_MODULE_NOT_FOUND` on `/app/services/home-relay/src/token.js` is gone: `/api/healthz` answers **HTTP 200** and `/api/home/status` and `/api/home/connections` answer 401 (auth required) rather than 500. What remains is ordinary staleness, 182 commits behind `main`, which is a routine cadence question and no longer a platform-health emergency. The 2026-09-04 text of this row, which described 96 commits and a 500ing healthz, is superseded. **Re-measured 2026-10-08 17:05 UTC:** production serves `fe2a8b24f` on `three-ws-api-00479-d7z`, **146 commits behind `main`, 90 of them fixes** (security, payouts, SSRF and XSS among them). Shipping them clears four live findings at once: the ring sweep's fee falls back to the treasury instead of the dry sponsor (`a9c996e72`, about 30 failed sweeps an hour since 2026-10-03), an EVM testnet identity-index cursor frozen since 2026-09-11 (`21fa42bb5`), two pages and their two docs pages advertised in the sitemap but answering 404 (added in `16f9d07f4`), and `/tty` holding its animation until the end for any client that accepts compression (`03b4f1189`). One migration must go with it: `20261008150000_revenue_direct_settle_backfill.sql` (pending; it matches 0 rows today, so applying it is safe, and the code half is what keeps the payout cron from paying a direct-to-seller sale twice). Commands, in order: `npm run db:status`, `npm run db:migrate`, `npm run deploy:gcp:full`. **Re-measured 2026-10-09 03:05 UTC by the triage sweep:** production still serves image `fe2a8b24f` (now on revision `three-ws-api-00480-mws`, an env-only roll that raised `DB_RETENTION_HIGH_WATER_MB` to 16384), **172 commits behind `main`**, and the `evolve` branch holds 85 more awaiting the runner's merge, six of them this sweep's production fixes: the leak scanner's lease and budget (`0abce6d69`, which ends about eight overlapping 600 to 900 s scans that were burning the Solana RPC quota), `-32020` RPC failover (`4b51a179c`), static 412/416 instead of 500 (`e01e43b3b`), the news knowledge write fix (`833407d4f`, 120 lost stories a day), and the reconstruct sweep's plan-limit close (`85d3b2034`). `20261008150000_revenue_direct_settle_backfill.sql` has since been applied; `npm run db:status` now lists two pending, `20261008190000_oauth_refresh_user_client_index.sql` and `20261008200000_mcp_client_analytics.sql` (rows 27 and 28). The 8 sitemap pages still answering 404 (`/grok`, `/grok-skill.md`, `/docs/grok`, `/docs/grok-bot`, `/dextools`, `/docs/dextools`, `/chart-companion`, `/docs/chart-companion`) are all deploy lag. | one yes | Shipping whatever has landed on `main` since 2026-09-08 | [01-ship-readiness.md](../901-production-100-01-ship-readiness.md) |
+| 2 | Fund the settle sponsor wallet `WwwuGbqHrwF5RG89KhUbmRWEvjnRH9k5kVM5p7T3WwW`. Healthz reports `x402_settle` down with `cause: sponsor_floor` (settle 5.9%, 3 of 51 paid attempts over 3 hours, 329 `no_solana_accept`) and the sniper worker out of capital (0.2307 SOL to refill its wallets, funding master holds 0.0250 SOL); no config change moves either. Re-read 2026-09-04 08:21 UTC against mainnet: the sponsor holds **0.001507661 SOL** (1,507,661 lamports), still below the settle floor, so the Solana accept stays withdrawn from every 402 challenge. Two months of re-measurement have only ever moved this number down; nothing on this machine can move it up. **Re-measured 2026-10-08 17:05 UTC:** the sponsor holds **651,828 lamports (0.000652 SOL)**, below rent-exempt (890,880), and healthz reads `x402_settle` down at 0.0% (0 of 32 paid attempts in 3 hours, 58 settles refused at the fee-wallet floor). Because the sponsor cannot co-sign, every 402 challenge advertises the Solana accept without `feePayer`, and the ring skips it: 151 `missing_fee_payer` skips in 24 hours on the MCP streaming health probe alone. Last successful settle: 2026-10-03 19:24 UTC. **Row 1's deploy alone restores the ring, no funding needed** (diagnosed 2026-10-08 by read-only simulation): the treasury pays its own sweep fee (5,001 lamports; it holds 0.060 SOL, about 11,800 sweeps) and moves its stranded 2.587 USDC back to the ring payer, which holds 341 atomics today, and ring settles are self-pay. What still needs this wallet: the sponsor-mode callers (the loop's inline entries and the MCP streaming health probe) and the facilitator's `sponsor_fee_unfunded` verify rejects. Clearing the 2,000,000-lamport floor takes at least **1,348,172 lamports**. **Re-measured 2026-10-09 03:05 UTC:** still **651,828 lamports**, unchanged; healthz reads `x402_settle` down at 0.0% (0 of 39 paid attempts in 3 hours, 37 `sweep_broadcast_failed`, 53 settles refused at the fee-wallet floor). The sweep failures are the ones row 1's deploy clears (`a9c996e72`). | ~1 SOL | x402 settle recovery, backlog 01 retirement, sniper | [../backlog/PROGRESS.md](backlog-PROGRESS.md), 2026-08-02 and 2026-09-01 entries |
| 3 | Decide the stranded customer funds: customers hold about 0.35 SOL behind the retired encryption key and cannot withdraw. Credit (recommended), contact first, or write off. The brief now exists with the measurement, the impossibility argument, the cost of each option and the exact commands: **[docs/ops/stranded-wallets.md](../../docs/ops/stranded-wallets.md)**. Measured 2026-09-02 from production's own records: the platform side is exactly two bot wallets holding 0.142875505 SOL (the reclaim cron has failed on them ~35,700 times each, most recently today), and one customer agent is confirmed by a `wallet_key_retired` withdraw failure (`GemVS5fT958FKRe5fpgizohUYUKE8cUDueEdmB1bmXnm`, 0.250001 SOL). **Both customer wallets are now named** (2026-09-04): the keyed audit was run against production, so this row no longer waits on anything. `GemVS5fT958FKRe5fpgizohUYUKE8cUDueEdmB1bmXnm` holds 0.250001 SOL and `HPL1LfuTdYDwtzJDzsnrmR2ngrrQwLTQyxJszCC4DHsN` holds 0.100001 SOL, 0.350002 SOL across exactly two wallet-auth accounts. Nine of 735 custodial wallets are sealed; the other five customer ones hold nothing. The brief carries the full table. Every input the decision needs is measured. | decision | Closing the support obligation | [docs/ops/stranded-wallets.md](../../docs/ops/stranded-wallets.md), [../backlog/PROGRESS.md](backlog-PROGRESS.md) |
| 4 | Rotate or explicitly accept three credentials, now that the storage half is done. **The config work shipped 2026-09-02**: the economy master key and 58 other plaintext credentials are Secret Manager references on `three-ws-api-00407-m7c`, zero plaintext credentials remain, healthz verified no worse and the master's signing path verified live. What is left is a judgment only the owner can make. (a) The master key was readable in the service config for some window, so decide whether to rotate it or accept that exposure, knowing who held project viewer access in that period; the exact owner-gated commands are in the runbook. (b) `WALLET_ENCRYPTION_KEY` and `JWT_SECRET` hold the SAME value on production, which passes the dedicated-key guard in `secret-box.js` while defeating what it guards; changing it is a re-key with real stranding risk, so it is not an agent's call. (c) `X402_SEED_SOLANA_SECRET_BASE58` holds the same value as `LAUNCHER_MASTER_SECRET_KEY_B64` though `scripts/wire-master-wallet.mjs` assigns those slots to different wallets; decide which is correct. | decision (+ one fund move if rotating) | 03 is otherwise closed | [docs/ops/wallet-key-migration.md](../../docs/ops/wallet-key-migration.md), 2026-09-02 section |
| 5 | Accept the Anthropic terms in Vertex Model Garden for `aerial-vehicle-466722-p5`, then an agent flips `VERTEX_CLAUDE_ENABLED=1` after a live probe. | minutes | First-party Claude lane | `docs/ops/llm-lanes.md` |
@@ -35,13 +35,14 @@ on it; some resolve themselves as other work lands.
| 21 | **Decide whether to rotate the lane credentials read on 2026-09-09.** Resolving the LLM chain for the fact-check benchmark printed secret material into an agent session transcript: the `vercel-inference@aerial-vehicle-466722-p5` service-account **private key** (`GCP_SERVICE_ACCOUNT_JSON`, key id `93be907a...`), plus `GROQ_API_KEY`, all five OpenRouter keys, `NVIDIA_API_KEY` and `OPENAI_API_KEY`. `INTERNAL_API_KEY` was used but never printed. Nothing indicates misuse and all of them remain valid; the exposure surface is whoever can read that transcript. Rotate or accept, and note that the OpenAI key is dead anyway (row 6) and OpenRouter key #1 is already revoked. | decision (+ minutes if rotating) | Closing out the exposure | [PROGRESS.md](production-100-PROGRESS.md), 2026-09-09 entry |
| 22 | **Decide whether to arm the three trading growth programs** (added 2026-09-29). Built, tested and disarmed: the big-win X poster (`BIG_WIN_X_ENABLED`, posts to @trythreews), the first-rug softener (`RUG_SOFTENER_ENABLED`, pays $THREE, default cap $25 per claim and $100 per day) and the early-leader bonus (`EARLY_LEADER_ENABLED`, pays $THREE, default 20 slots at $50 and $250 per week). Each is armed on its own. Steps: (1) `npm run db:status`, then `npm run db:migrate` for `20260929120000_growth_program_payouts.sql` (payout programs only); (2) set `THREE_PRIZE_PAYOUT_KEY` (unset on production as of 2026-09-29; the programs deliberately refuse the club treasury fallback, which is the economy master wallet) to a dedicated wallet holding the $THREE to cover the caps (payout programs only); (3) deploy; (4) read the dry run at `/api/ops/growth-programs`; (5) `gcloud run services update three-ws-api --region us-central1 --update-env-vars <FLAG>=1`. Measured 2026-09-29 against production data: nothing would be posted or paid today (no live win cleared the 0.1 SOL size gate, no first trade met the softener bar, no leader has a profitable 30-day record with followers). | one yes per program | Roadmap 917 items 3.8, 12.15, 13.4 | [docs/growth-programs.md](../../docs/growth-programs.md) |
| 23 | **Name the two $THREE receiver wallets** (added 2026-09-30). `/api/healthz` reports `three_token_rail` down and signed-in `POST /api/token/quote` answers `503 treasury_unavailable`, so no $THREE-priced purchase can be priced (Forge paid tiers, spins, scarcity mints, token marketplace sales, copy-trading fees). The quote signing key is done (`three-quote-secret`, wired on `three-ws-api-00466-85g`); only `THREE_TREASURY_WALLET` and `THREE_REWARDS_WALLET` remain. No revision ever held them, so there is nothing to restore. Candidates: treasury `wwwwwDxFWRn7grgr3Esrsg5C6NvDoDHSA4gaCffccrU` (x402 receiver, already holds a $THREE account), rewards `WwwuGbqHrwF5RG89KhUbmRWEvjnRH9k5kVM5p7T3WwW` (the economy master, which `REWARDS_DISTRIBUTOR_SECRET` signs as). The rewards candidate is also the buyback signer, so arming `THREE_BUYBACK_ENABLED` or `THREE_MICROBUY_ENABLED` would sweep the whole rewards pool into the treasury; decide that before arming either lane. Setting the two vars moves no funds by itself: `gcloud run services update three-ws-api --region us-central1 --update-env-vars THREE_TREASURY_WALLET=<treasury>,THREE_REWARDS_WALLET=<rewards>`. | two addresses | The whole $THREE purchase rail | [docs/money-map.md](../../docs/money-map.md#three-split-receivers-inbound-three) |
-| 24 | **Top up the paid Solana RPC plans** (added 2026-10-08). Helius answers `429: max usage reached` on every call: 45,461 such lines on `three-ws-api` in the 24 hours to 2026-10-08 17:10 UTC, the oldest retained one from 2026-09-08. healthz reads `helius` degraded and `rpc_lanes` "all 2 paid lanes exhausted", so Solana traffic runs on throttled free nodes. That is the whole cause of `agent_index` reading down: 172 of the 176 erroring Solana cursors carry `solana rpc 429 @ public.rpc.solanavibestation.com`, the last free lane, and the rest are a free provider's own errors. No config change adds credits, and promoting the reserve lane is the documented wrong fix. Upgrade or top up the Helius plan (and the second paid lane), then confirm `helius` and `rpc_lanes` read ok on `curl -s https://three.ws/api/healthz`. | billing | Solana agent index freshness, balance reads, every on-chain route's latency | [docs/ops/production-log-triage.md](../../docs/ops/production-log-triage.md), "Every paid Solana RPC lane exhausted at once" |
+| 24 | **Top up the paid Solana RPC plans** (added 2026-10-08). Helius answers `429: max usage reached` on every call: 45,461 such lines on `three-ws-api` in the 24 hours to 2026-10-08 17:10 UTC, the oldest retained one from 2026-09-08. healthz reads `helius` degraded and `rpc_lanes` "all 2 paid lanes exhausted", so Solana traffic runs on throttled free nodes. That is the whole cause of `agent_index` reading down: 172 of the 176 erroring Solana cursors carry `solana rpc 429 @ public.rpc.solanavibestation.com`, the last free lane, and the rest are a free provider's own errors. No config change adds credits, and promoting the reserve lane is the documented wrong fix. Upgrade or top up the Helius plan (and the second paid lane), then confirm `helius` and `rpc_lanes` read ok on `curl -s https://three.ws/api/healthz`. **Re-measured 2026-10-09 03:05 UTC:** `rpc_lanes` still reads "all 2 paid lanes exhausted", and `agent_index` reads down at a 36m median Solana lag with 135 of 1,663 cursors erroring. One cause of the burn was ours and is fixed on `evolve` pending row 1's deploy: the ring leak scanner ran about eight overlapping copies at all times (213 runs in 6 hours, average 788 s, each walking 2,006 registry wallets), now one leased, 200 s-budgeted run per 10 minutes (`0abce6d69`). That cuts the platform's own RPC demand; it does not add credits, so this row stands. | billing | Solana agent index freshness, balance reads, every on-chain route's latency | [docs/ops/production-log-triage.md](../../docs/ops/production-log-triage.md), "Every paid Solana RPC lane exhausted at once" |
| 25 | **Deploy so cloud MCP connectors can sign in to four hosted servers** (added 2026-10-08 by the evolve loop, x-grok order 023). Production (`fe2a8b24f`, revision `three-ws-api-00479-d7z`) still points every OAuth 401 at the root protected-resource document naming `/api/mcp`, so a stock MCP SDK client (Grok Bot, claude.ai connectors) refuses to start sign-in on `/api/mcp-3d`, `/api/mcp-agent`, `/api/mcp-bazaar` and `/api/ibm-mcp`. The fix is on `main` (`7985f1cce`, `59bdf5707`) and the probe passes all seven servers against local code. The submit gate also needs the pending `20261008150000_revenue_direct_settle_backfill.sql` applied first (`npm run db:status`, then `npm run db:migrate`), which is another session's payout backfill, so read it before applying. Then `npm run deploy:gcp:full`, then `npm run probe:mcp-clients -- --json prompts/x-grok/_generated/connector-probe.json`, which must print `all 7 servers passed`. | one deploy | Grok Bot and every OAuth connector on four servers; closes x-grok order 023 | [docs/mcp.md](../../../docs/mcp.md#client-compatibility), [prompts/x-grok/_generated/connector-probe.json](../../x-grok/_generated/connector-probe.json) |
| 26 | **Approve committing the regenerated MCP catalog** (added 2026-10-08 by the evolve loop, x-grok order 024 and order 401). `npm test` is green on the evolve working tree (vitest 2194 files / 31,910 tests, Playwright 262 passed with `E2E_PORT=3107`), but three files stay uncommitted under the coin commit gate because their regenerated text names other crypto projects already named in committed source: `public/mcp-catalog.json` (fills the missing input schemas of `agora_quote_register` and `preview_agent_identity` and adds `pay_quote`; without it `tests/mcp-schema.test.js` fails 3 tests at HEAD), `tests/fixtures/mcp-golden-tools.json` (the `audit:mcp-golden` contract refresh) and a one-line dash fix in `packages/agora-mcp/src/tools/register.js`. They are held as an `evolve held:` stash. To approve, on `main`: `git stash list | grep 'evolve held'`, review with `git stash show -p <ref>`, then `git stash apply <ref>` and commit those three paths; or regenerate instead with `npm run build:mcp-catalog && node scripts/audit-mcp-golden.mjs --update` and commit `public/mcp-catalog.json tests/fixtures/mcp-golden-tools.json`. Then `E2E_PORT=3107 npm test` should exit 0 at HEAD and orders 024 and 401 can retire. | one yes | Orders 024 and 401 | [x-grok-PROGRESS.md](x-grok-PROGRESS.md), 2026-10-08 order 024 retry entry |
| 27 | **Apply the OAuth refresh-token index, then ship instant revocation with the next deploy** (added 2026-10-08 by the evolve loop, x-grok order 025, which is otherwise complete and retired). `cf30be269` makes every OAuth access token check its grant on each request, so Revoke in Settings, Connected apps cuts a cloud connector such as Grok Bot off on its next call instead of up to an hour later. That lookup filters `oauth_refresh_tokens` on `(user_id, client_id)` including revoked rows, which the existing partial index cannot serve; at 4,228 rows today it is fast without one, but the table gains a row per refresh. The code is correct without the index. Apply it with `node scripts/apply-migrations.mjs --apply --file 20261008190000_oauth_refresh_user_client_index.sql` (or `npm run db:status`, then `npm run db:migrate`, which also applies row 25's backfill, so read that one first). It is pending until then and `npm run db:check` blocks a deploy on it. The behavior itself ships with the deploy in row 25. | one command | Fast grant checks on every OAuth MCP call; the deploy gate | [docs/mcp.md](../../../docs/mcp.md#cloud-connectors-and-revocation) |
| 28 | **Apply the MCP client analytics migration, then ship it with the next deploy** (added 2026-10-08 by the evolve loop, x-grok order 030, which is otherwise complete and retired). `e8381ce9c` makes every hosted MCP server issue an `Mcp-Session-Id` on initialize and record which AI client (Claude, ChatGPT, Cursor, Grok Bot, ...) opened each session and which tools it calls, read by the admin board at `/mcp-clients`. The tables do not exist in production yet: `npm run db:status` on 2026-10-08 lists `20261008200000_mcp_client_analytics.sql` pending beside rows 25 and 27's migrations. Until it is applied the servers keep working (each flush logs `[mcp-clients] session write failed` and drops the batch) and the board shows "The analytics tables are not there yet". Apply it alone with `node scripts/apply-migrations.mjs --apply --file 20261008200000_mcp_client_analytics.sql`, or with the rest via `npm run db:status` then `npm run db:migrate` (read row 25's backfill first). `npm run db:check` blocks the deploy on it until then. One more change is held for you as an evolve stash (`git stash list`, "evolve held"), not committed: the same hook on the eighth server in `/.well-known/mcp.json`, held only by the coin commit gate because that server is built around another project; approve it and that surface is counted too (it was verified locally with the rest). | one command, one approval | Per-client MCP usage on `/mcp-clients` | [docs/ops/mcp-clients.md](../../../docs/ops/mcp-clients.md) |
| 29 | **After the row 25 deploy, rerun the Grok Bot doc samples against production** (added 2026-10-09 by the evolve loop, x-grok order 035, which is otherwise complete and retired). `docs/grok-bot.md` (`ae20babcd`) documents `/api/mcp-grok`, install tokens, `get_job` and `idempotency_key`, none of which production (`fe2a8b24f`) serves yet, so its ten samples were run against a server built from this tree (all passed, `npm run probe:mcp-clients -- --only mcp-grok` included) and against production, where they answer 404. Once the deploy lands, run `set -a; . ./.env; set +a; node scripts/run-doc-samples.mjs docs/grok-bot.md` and expect `0 of 10 samples failed` with real JSON in every output (it makes one free `forge_free` generation). Nothing to approve beyond the deploy itself. | one command after the deploy | Proof that every Grok Bot doc sample runs on production as written | [docs/grok-bot.md](../../../docs/grok-bot.md#keeping-this-page-true) |
| 30 | **After the row 25 deploy, rerun `npx three-ws setup --client grok-bot` against production** (added 2026-10-09 by the evolve loop, x-grok order 037, which is otherwise complete and retired). The CLI (`cd9b7f5be`) reads the live directory, so on production today (`fe2a8b24f`) it connects the free studio `three-ws-studio` and its live check passes; `--auth install` answers that the deployment cannot mint install tokens yet, and `--auth key` without `--connector-key` answers that it cannot mint connector keys from the CLI yet, because the bearer `preset: "connector"` on `POST /api/api-keys` (`3350e513c`) is not deployed. All four modes were verified against a server built from this tree, including a real connector-key mint (scope capped to the connector set, then revoked). Once the deploy lands, run `node packages/three-ws-cli/src/cli.js setup --client grok-bot --yes`, then the same with `--auth install` and `--auth oauth`; each must name `three-ws-grok` and exit 0. Nothing to approve beyond the deploy itself. | three commands after the deploy | Proof the Grok Bot CLI setup reaches `/api/mcp-grok`, install tokens and connector keys on production | [docs/cli.md](../../../docs/cli.md#grok-bot) |
+| 31 | **Redeploy the GPU workers that fetch caller images, so a Wikimedia reference photo stops failing** (added 2026-10-09 by the evolve loop triage sweep). `upload.wikimedia.org` answers httpx's default `python-httpx` agent with 403, so `model-triposg` failed every job whose reference photo was a Wikimedia link before inference (4 logged 403 failures in the 24 hours to 2026-10-09 03:00 UTC). `d886d09e1` sends a descriptive User-Agent from the shared `worker_security.py`, mirrored to all 17 vendored copies, but a worker only picks it up when its image is rebuilt. Start with the one that failed: `gcloud builds submit --config workers/model-triposg/cloudbuild.yaml --region us-central1 --project aerial-vehicle-466722-p5 --substitutions=SHORT_SHA=manual$(date +%s) .` (the command in `workers/model-triposg/README.md`); the other image-fetching workers (`model-trellis`, `model-hunyuan3d`, `rig`, `texture`, `stylize`, `segment`, `remesh`) inherit it on their next routine deploy. Separately worth a look: `model-trellis` was killed by its own liveness probe (`/health` timing out during inference, then signal 9) three times in the 24 hours to 2026-10-09, which left 147 task polls answering 503; not root-caused this session. | one build | Image-to-3D jobs whose reference photo is a Wikimedia link | [workers/model-triposg/README.md](../../../workers/model-triposg/README.md) |
Deleted 2026-09-01: the former row 5 (`WALLET_ENCRYPTION_KEY_PREVIOUS` policy), which
declared itself "delete on reading" and had outlived that by three weeks.