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): record the 2026-09-29 agent-sniper deploy and the first fleet shadow review in order 918 and owner row 19

nirholas4ec130f48161
diff --git a/prompts/finish/918-roadmap-pumpfun-trading-arena.md b/prompts/finish/918-roadmap-pumpfun-trading-arena.md
index c0c82c03f..8dca21529 100644
--- a/prompts/finish/918-roadmap-pumpfun-trading-arena.md
+++ b/prompts/finish/918-roadmap-pumpfun-trading-arena.md
@@ -1,6 +1,6 @@
# The Pump.fun Trading Arena — Master Plan
-**Status:** Strategy / roadmap. Owner-facing. Last updated 2026-09-24.
+**Status:** Strategy / roadmap. Owner-facing. Last updated 2026-09-29.
**Scope:** ONLY pump.fun trading, deploying, agent monetization, and copy-trading. Nothing else.
> **Build status (verified 2026-09-03).** Phases 0 to 6 are shipped, on surfaces that
@@ -17,10 +17,10 @@
> Section 4's anti-gaming layer was the last real gap and closed on 2026-09-03
> (`api/_lib/copy-eligibility.js`): the copyable-status sybil bar (4.4), the follower
> drawdown circuit breaker (4.5), and self-copy / self-follow exclusion (4.6).
-> Section 9's named blocker, AMM exits on graduated positions, is written and
-> committed (`workers/agent-sniper/amm-exit.js`, imported at HEAD by
-> `executor.js`, `positions.js` and `graduation-ride.js`). It is not yet running:
-> see the deploy paragraph below.
+> Section 9's named blocker, AMM exits on graduated positions, is written,
+> committed and running in production (`workers/agent-sniper/amm-exit.js`,
+> imported by `executor.js`, `positions.js` and `graduation-ride.js`; the fleet
+> rolled onto HEAD on 2026-09-29, see the owner paragraph below).
>
> **Prompt D, the adversarial Risk Officer (section 6), is now BUILT**
> (`workers/agent-sniper/risk-officer.js`, migration
@@ -40,63 +40,73 @@
> positions that actually opened, so their realized P&L is the evidence for or
> against arming it (the query is in `workers/agent-sniper/README.md`).
>
-> **Owner: two owner-gated steps remain, both by design, and the first one is
-> larger than the Risk Officer.** (1) Deploy the agent-sniper worker. Its running
-> 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 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
-> exists, arm enforcement with `SNIPER_RISK_OFFICER=enforce` fleet-wide (via
-> `--update-env-vars`, never `--set-env-vars`) or per strategy via
-> `risk_officer_level` (gate 1: it changes what real SOL buys). Both are tracked
-> as row 19 of
+> **Owner: one owner-gated step remains, plus a capital top-up. Step (1), the
+> agent-sniper deploy, is DONE (2026-09-29, owner approval "finish them all").**
+> Measured, not assumed:
+>
+> - **What the old image lacked.** Revision `agent-sniper-00033-rg6` ran image
+> `sha256:9946506e…`, Cloud Build `1deea978` from a working-tree upload of
+> 2026-08-09 09:47 UTC (the revision itself was rolled 2026-08-11). Since that
+> upload, 15 commits touched `workers/agent-sniper/` (9 files, +996/-66):
+> `risk-officer.js` is new, and the executor, config, journal, intel learner,
+> reconcile and radar changed. `amm-exit.js` was NOT missing, contrary to the
+> earlier wording here: it has been in the image since June, and only its SDK-2
+> pool fields (`quoteMint`, `isMayhemMode`, `creatorFeeBps`) were behind.
+> - **The first attempt was refused, and that exposed a real bug.** Revision
+> `agent-sniper-00034-7zp` crashed on boot with `ENOENT /app/data/plans.json`:
+> since 2026-09-25 `api/payments/_config.js` reads that file at load, the sniper
+> reaches it through the trade-fee module, and the Dockerfile never copied
+> `data/`. Cloud Run kept 100% on 00033, so the fleet never stopped. Fixed in
+> `b17456355` (the Dockerfile copies `data/plans.json`, and `npm run
+> deploy:sniper` now boots `index.js` from exactly the Dockerfile COPY set
+> before it spends a Cloud Build, which reproduces that crash in 10 seconds).
+> - **What runs now.** `agent-sniper-00035-krc`, image `sha256:5b115be9…` built
+> from commit `b17456355`, serving 100%, worker booted 07:07:13 UTC. The env is
+> byte-identical to the old revision (26 keys, `SNIPER_MODE=live`,
+> `SNIPER_RISK_OFFICER` unset so the default is `shadow`), and the old instance
+> drained cleanly ("bye" at 07:07:21). All 31 sniper-related test files
+> (504 tests) passed at that commit first. No new ERROR-severity logs on 00035;
+> the only error-level line is the pre-existing "funding master cannot cover
+> refills".
+> - **Shadow rows exist from the fleet.** In the 32 minutes watched after the
+> boot the fleet opened one live position (07:14:35 UTC, `oracle_crossing`,
+> 0.002 SOL), and `sniper_risk_reviews` recorded its shadow review 1.6 seconds
+> later (severity `none`, 1,079 ms, answered by the platform chain's
+> `nvidia/nemotron-3-super-120b-a12b` rung), joined by `position_id`. That
+> position then exited through the new image at 07:30:01 UTC
+> (`liquidity_decay`, realized -0.000355 SOL). The table now holds 8 rows: the
+> 7 from the 2026-09-04 local verification plus the first fleet row.
+>
+> **What remains, with owners.** (2) Arm the Risk Officer to `enforce` (gate 1:
+> it changes what real SOL buys), fleet-wide with `SNIPER_RISK_OFFICER=enforce`
+> via `--update-env-vars` (never `--set-env-vars`) or per strategy via
+> `risk_officer_level`, and only once the shadow-P&L query in
+> `workers/agent-sniper/README.md` has enough closed, joined rows to show the
+> `block` (and `caution`) severities losing more than `none`. One row is not evidence; at
+> the current rate that takes weeks. (3) An owner spend, not a code task: the
+> fleet is capital-starved. `/api/healthz` reads "7/11 armed wallets can still
+> trade (4 starved, 1 shrunk). Refilling them needs 0.2397 SOL. The funding
+> master holds only 0.0250 SOL", and the buy rate (26 opens in the 24 hours
+> before the deploy) and therefore the shadow-evidence rate scale with that
+> refill. Both are row 19 of
> [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.
+> Everything else in this plan is written, tested, deployed and applied to the
+> production schema.
>
-> **Re-verified 2026-09-24.** Still exactly two owner-gated steps, nothing else.
-> Against the production database at 06:50 UTC: 23 fleet buys in the previous 24
-> hours (last at 06:46 UTC), `sniper_risk_reviews` still at the same 7 rows from
-> 2026-09-04 with none from the fleet, and all 28 strategy rows (13 enabled) at
-> `risk_officer_level = shadow`. So the agent-sniper image still predates
-> `risk-officer.js` and `amm-exit.js`; `npm run deploy:sniper` remains step (1).
-> One product gap in section 3.2 closed the same day: "why it traded" was only
+> One product gap in section 3.2 closed on 2026-09-24: "why it traded" was only
> visible to the owner (the private `/api/sniper/journal`). Every public trade
> now carries a receipt of the evidence its gates recorded before the entry
> (`api/_lib/trade-receipt.js`, `GET /api/sniper/receipt`, the "why" drawer on
> `/trader/:id`, the "Why it traded" section on `/trade/:id`, the `trade_receipt`
> MCP tool; doc `docs/trade-receipts.md`).
>
-> **Re-verified 2026-09-25, unchanged.** Production database at 05:14 UTC: 23
-> fleet buys in the previous 24 hours (last at 05:12 UTC), zero
-> `sniper_risk_reviews` rows in that window (still the same 7 from 2026-09-04),
-> and all 28 strategy rows (11 enabled) at `risk_officer_level = shadow`. gcloud
-> was logged in this time and confirms it directly: the service still serves
-> `agent-sniper-00033-rg6`, built 2026-08-11, with 14 commits to
-> `workers/agent-sniper/` since. `node scripts/deploy-sniper.mjs --dry-run` is
-> clean (runtime SA, registry and all required secrets present; it rolls
-> `services update --image` and preserves the live env), and the 28 sniper test
-> files (468 tests, including `agent-sniper-amm-exit`, `sniper-risk-officer` and
-> `copy-eligibility`) pass at HEAD. Step (1) is still one command waiting on the
-> owner's yes.
->
-> **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.
+> **Prove the deploy state from the database, not only from `gcloud`.** 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 mean the running image predates `risk-officer.js`. Before
+> 2026-09-29 that was 41 buys against 0 fleet rows; after it, every new
+> position carries a review row, which is how a future regression to an old
+> image shows up.
**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 55953eb59..53fba947a 100644
--- a/prompts/finish/_context/production-100-OWNER-ACTIONS.md
+++ b/prompts/finish/_context/production-100-OWNER-ACTIONS.md
@@ -9,7 +9,7 @@ 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 (2026-09-09: 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 (2026-09-29: 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 |
|---|---|---|---|---|
@@ -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; 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) |
+| 19 | Arm the agent-sniper Risk Officer once shadow evidence justifies it, and refill the sniper fleet's trading capital. **The deploy half of this row is DONE (2026-09-29, owner approval "finish them all").** Measured: the old revision `agent-sniper-00033-rg6` ran an image uploaded 2026-08-09 09:47 UTC, 15 commits behind on `workers/agent-sniper/` (9 files, +996/-66, incl. the new `risk-officer.js`; `amm-exit.js` was already in it, only its SDK-2 pool fields were behind). The first roll (`agent-sniper-00034-7zp`) was refused by Cloud Run because the image lacked `data/plans.json`, read at load by `api/payments/_config.js` since 2026-09-25; traffic never left 00033. Fixed in `b17456355`, which also makes `npm run deploy:sniper` boot the worker from the exact Dockerfile COPY set before building. `agent-sniper-00035-krc` (commit `b17456355`) now serves 100% with a byte-identical env (`SNIPER_MODE=live`, `SNIPER_RISK_OFFICER` unset, so `shadow`), no new ERROR logs, and the fleet trading through it: one live buy at 07:14:35 UTC got its shadow review 1.6 s later (severity `none`, joined by `position_id`), and that position exited at 07:30:01 UTC (`liquidity_decay`). `sniper_risk_reviews` now has 8 rows, the first 1 from the fleet. **Remaining, both owner-only:** (a) arming `enforce` is gate 1 (it changes what real SOL buys): `SNIPER_RISK_OFFICER=enforce` via `--update-env-vars` (never `--set-env-vars`) or per strategy via `risk_officer_level`, only after the shadow-P&L query in `workers/agent-sniper/README.md` shows `block`/`caution` calls losing more than `none` across enough closed positions (one row is not evidence). (b) a SOL spend: `/api/healthz` reports 7/11 armed wallets tradeable (4 starved, 1 shrunk), 0.2397 SOL needed, funding master holds 0.0250 SOL; the buy rate, and so the evidence rate for (a), scales with that refill. | one spend now, one decision later | The rate of shadow evidence for the Risk Officer, and 4 starved fleet wallets | [../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 | **Re-confirmed 2026-09-10, and the blast radius is wider than this row said.** The denial covers the Generative Language API (`generativelanguage.googleapis.com`, the `GOOGLE_API_KEY` path) as well as Vertex, so there is no Google lane of any kind. It is also the real blocker behind **order 311's i18n translation**: that order names `gcloud auth login` as its owner action, but gcloud auth WORKS in this workspace, and the actual constraint is that every reachable free lane is throttled per key (NVIDIA about 27 keys/min per process and `429`s under parallelism, Groq capped at 200k tokens/day, OpenAI `billing_not_active`). Vertex is the only lane billed to the credit pool with no free-tier quota, so it is the only one that finishes 84 locales in one pass. Clearing this hold therefore unblocks **three** orders at once: 903, 907 and 311.
| 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) |