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(sniper-arena): prove the agent-sniper deploy state from the database, not gcloud

nirholas420be782419e
diff --git a/prompts/finish/918-roadmap-pumpfun-trading-arena.md b/prompts/finish/918-roadmap-pumpfun-trading-arena.md
index a29d28575..471b61d26 100644
--- a/prompts/finish/918-roadmap-pumpfun-trading-arena.md
+++ b/prompts/finish/918-roadmap-pumpfun-trading-arena.md
@@ -45,10 +45,11 @@
> image was built 2026-08-11, while 48 files under `workers/agent-sniper/` have
> been committed since, so the fleet trading live right now has neither
> `amm-exit.js` (section 9's blocker) nor `risk-officer.js`. Re-measured
-> 2026-09-09 against the production database: the worker is alive and buying
-> (heartbeat current, positions opened the same afternoon), every strategy row
-> reads `risk_officer_level = shadow`, and `sniper_risk_reviews` still holds only
-> the 7 rows from the 2026-09-04 local verification, none of them from the fleet.
+> 2026-09-09 19:34 UTC against the production database: the worker is alive and
+> buying (41 positions opened in the previous 24 hours, the last at 19:02 UTC),
+> all 27 strategy rows (13 enabled) read `risk_officer_level = shadow`, and
+> `sniper_risk_reviews` still holds only the 7 rows from the 2026-09-04 local
+> verification, none of them from the fleet.
> The deploy is one command, `npm run deploy:sniper`, and on an existing service
> it rolls `gcloud run services update --image`, which preserves the running env
> and does not demote the fleet to simulate (gate 2). (2) Once shadow evidence
@@ -59,6 +60,17 @@
> [production-100-OWNER-ACTIONS.md](_context/production-100-OWNER-ACTIONS.md).
> Everything else in this plan is written, tested and applied to the production
> schema.
+>
+> **Prove the deploy state from the database, not from `gcloud`.** The Codespace
+> loses its gcloud login on every recycle, so `gcloud run services describe
+> agent-sniper` (the image-date check) is usually unavailable, and its failure
+> says nothing about the fleet. The database answers the same question and is
+> always reachable: a shadow review is recorded fire-and-forget on EVERY buy
+> (`reviewBuy` in `executor.js`, `insert into sniper_risk_reviews` in
+> `risk-officer.js`), so fleet buys with no matching review rows can only mean
+> the running image predates `risk-officer.js`. 41 buys against 0 fleet review
+> rows is that proof. After the deploy, the same counts turning non-zero are how
+> you know it took.
**The only coin this platform promotes is `$THREE` (`FeMbDoX7R1Psc4GEcvJdsbNbZA3bfztcyDCatJVJpump`).** Pump.fun coins traded/launched through the platform are user runtime data, never endorsements.
---
diff --git a/prompts/finish/_context/production-100-OWNER-ACTIONS.md b/prompts/finish/_context/production-100-OWNER-ACTIONS.md
index 364e79ee1..c03a2bcad 100644
--- a/prompts/finish/_context/production-100-OWNER-ACTIONS.md
+++ b/prompts/finish/_context/production-100-OWNER-ACTIONS.md
@@ -30,7 +30,7 @@ on it; some resolve themselves as other work lands.
| 15 | Re-authenticate `gcloud` in this workspace (`gcloud auth login`) **only if it stops answering**. This row is a standing warning, not a standing blocker, and 2026-09-04 is the clearest proof yet: `gcloud` answered every call this session with no intervention, and the backlog it had been holding was drained in one sitting. Cleared by that session: row 18 (`MULTIPLAYER_INTERNAL_URL` set, `/api/play/population` now `ok:true`), row 3's keyed audit (both stranded customer wallets named), and the event closeout's log read (`textPayload:"souvenir laurel-meetup"` over 30d returned **zero** lines against a control proving the stream is intact, corroborating the zero-grant finding before the window ages out). **Retry a real call before treating this row as a blocker, and never park work on it.** One dependent genuinely remains and is NOT a gcloud problem: the durable Upstash `player:*` scan needs to execute inside the VPC (the store is at a private `10.128.15.228` SRH proxy, unreachable from this Codespace), and creating the Cloud Run job to do that was refused by the tool policy here. That scan is confirmation only; the log read plus the code argument already settle it. Still gated on a live token when it lapses: the Cloud Scheduler comparison (fix-queue 03) and the fleet checks in quality-bar 03. | minutes, when it lapses | fix-queue 03, quality-bar 03 | CLAUDE.md self-unblock playbook, [PROGRESS.md](event-PROGRESS.md) |
| 16 | Create the `three-ws` GitHub organization and `three-ws/examples` (public), then run the push command `npm run export:satellites` prints. The export now stages 72 files and passes all four smoke stages online (registry, install, links); re-probed 2026-09-03, both the organization and the repository still answer 404. | minutes | roadmap developer-resources retirement | [docs/ops/examples-repo-export.md](../../docs/ops/examples-repo-export.md) |
| 17 | Apple Developer account. The WidgetKit extension, the Mac menu bar app and the iOS target are written and guarded (`npm run check:apple-widget`); the account is what signs them, for a Developer ID Mac build and an App Store iOS build alike. Separately, publish the Android 1.1.0 build (versionCode 2) to Play. | account + minutes | roadmap native-widgets task 4 and its release | [../roadmap/native-widgets.md](../916-roadmap-native-widgets.md) |
-| 19 | Deploy the `agent-sniper` worker, then decide separately whether to arm the Risk Officer. Measured 2026-09-09 against Cloud Run and mainnet: the service runs revision `agent-sniper-00033-rg6`, whose image was built **2026-08-11**, while it trades `SNIPER_MODE=live` right now (heartbeat 6s old, 11 strategies, 3,482 positions, one opened 05:10 UTC that morning). **48 files and about 11,000 lines under `workers/agent-sniper/` are committed and not in that image**, including `amm-exit.js`, which is the roadmap's own named blocker: `executor.js`, `positions.js` and `graduation-ride.js` import it at HEAD, so the running image has no code path that exits a position after its coin graduates. The Risk Officer (`risk-officer.js`, migration already applied in production) has therefore never run in production either: `sniper_risk_reviews` holds 7 rows, all from the 2026-09-04 local verification, and the shadow-evidence query in `workers/agent-sniper/README.md` runs green against the live schema and returns zero rows. The deploy is one command, `npm run deploy:sniper`, dry-run clean on 2026-09-09 (gcloud authed, runtime SA and Artifact Registry repo present, both required secrets present); on an existing service it rolls `gcloud run services update --image`, which preserves the running env, so it does **not** demote the fleet to simulate. Only the optional `telegram-alerts-chat-id` secret is missing, which disables Telegram ops alerting. Arming enforcement is a **separate, later** yes: `SNIPER_RISK_OFFICER=enforce` fleet-wide (via `--update-env-vars`, never `--set-env-vars`) or per strategy via `risk_officer_level`, and it should wait until shadow rows exist to justify it. | one yes now, one decision later | The trading Arena copy-trade loop (graduated-position exits), and all shadow evidence for the Risk Officer | [../918-roadmap-pumpfun-trading-arena.md](../918-roadmap-pumpfun-trading-arena.md), [workers/agent-sniper/README.md](../../workers/agent-sniper/README.md) |
+| 19 | Deploy the `agent-sniper` worker, then decide separately whether to arm the Risk Officer. Measured 2026-09-09 against Cloud Run and mainnet: the service runs revision `agent-sniper-00033-rg6`, whose image was built **2026-08-11**, while it trades `SNIPER_MODE=live` right now (heartbeat 6s old, 11 strategies, 3,482 positions, one opened 05:10 UTC that morning; re-measured 19:34 UTC the same day: 27 strategies with 13 enabled, 3,511 positions, 41 of them opened in the previous 24 hours and the last at 19:02 UTC). **48 files and about 11,000 lines under `workers/agent-sniper/` are committed and not in that image**, including `amm-exit.js`, which is the roadmap's own named blocker: `executor.js`, `positions.js` and `graduation-ride.js` import it at HEAD, so the running image has no code path that exits a position after its coin graduates. The Risk Officer (`risk-officer.js`, migration already applied in production) has therefore never run in production either: `sniper_risk_reviews` holds 7 rows, all from the 2026-09-04 local verification, and the shadow-evidence query in `workers/agent-sniper/README.md` runs green against the live schema and returns zero rows. **Re-verify this from the database, not from `gcloud`:** the Codespace loses its gcloud login on every recycle, so `gcloud run services describe agent-sniper` is usually unavailable and its failure says nothing about the fleet. A shadow review is recorded fire-and-forget on every buy, so fleet buys with no matching review rows can only mean the running image predates `risk-officer.js`; 41 buys against 0 fleet review rows is that proof, and those counts turning non-zero is how you confirm the deploy took. The deploy is one command, `npm run deploy:sniper`, dry-run clean on 2026-09-09 (gcloud authed, runtime SA and Artifact Registry repo present, both required secrets present); on an existing service it rolls `gcloud run services update --image`, which preserves the running env, so it does **not** demote the fleet to simulate. Only the optional `telegram-alerts-chat-id` secret is missing, which disables Telegram ops alerting. Arming enforcement is a **separate, later** yes: `SNIPER_RISK_OFFICER=enforce` fleet-wide (via `--update-env-vars`, never `--set-env-vars`) or per strategy via `risk_officer_level`, and it should wait until shadow rows exist to justify it. | one yes now, one decision later | The trading Arena copy-trade loop (graduated-position exits), and all shadow evidence for the Risk Officer | [../918-roadmap-pumpfun-trading-arena.md](../918-roadmap-pumpfun-trading-arena.md), [workers/agent-sniper/README.md](../../workers/agent-sniper/README.md) |
| 20 | **Clear the GCP billing hold on `aerial-vehicle-466722-p5`.** Every Vertex surface answers `403 Lightning dunning decision is deny for project: projects/93741856042`, confirmed 2026-09-09 both locally and in production's own logs (Vertex Imagen falling back at 05:33 UTC). This is not IAM and nothing in code routes around it: it takes out the Vertex Gemini anchor, Imagen, and `groundedSearch`, which is the fact-checker's ONLY working search rung (no Brave/Tavily/Exa/Serper key exists). With it down the chain falls to Wikipedia alone: a benchmark run that day saw 190 sources, 100% Wikipedia, and 15 of 38 claims had zero stance-bearing evidence. It is a billing/payment state on the account, so only the owner can clear it. | minutes, at the billing console | Grounded search, Vertex Gemini, Imagen, and publishing the fact-check accuracy run (04b) | [PROGRESS.md](production-100-PROGRESS.md), 2026-09-09 entry |
| 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 |