Borrowing it
Nothing to install: this file belongs to oxy-hq/oxygen. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/oxy-hq/oxygen/main/.claude/skills/oxy-scaling-design/SKILL.mdgit clone --depth 1 https://github.com/oxy-hq/oxygenWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/oxy-hq/oxygen/oxy-scaling-design)<a href="https://agentmods.dev/skills/oxy-hq/oxygen/oxy-scaling-design"><img src="https://agentmods.dev/badge/skills/oxy-hq/oxygen/oxy-scaling-design/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/oxy-hq/oxygen/oxy-scaling-design"><img src="https://agentmods.dev/badge/skills/oxy-hq/oxygen/oxy-scaling-design.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00156 | $0.02382 |
| Opus 5.5 | $0.00062 | $0.00953 |
| Sonnet 5.5 | $0.00031 | $0.00476 |
| Haiku 4.5 | $0.00016 | $0.00238 |
Grade A, and why
oxy-scaling-design scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 76 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Oxy multi-instance scaling — quick reference
The current architecture in one line: Oxy runs as a split fleet keyed by OXY_ROLE — a single stateful ide (the Factory: working copy + .git + compile + chat-run execution), a horizontally-scaled stateless serve fleet (reads Postgres + S3 only), and a worker queue drainer. This is a Postgres primary/read-replica shape: the ide is the primary, the serve fleet is the read replicas, and the compile boundary (compiled *_definitions rows + S3 blobs, keyed by revision_id) is the replicated read model.
The direction (decided 2026-10-05): the Factory is being retired, not scaled. No IDE; content is authored in git and pushed; compile becomes its own service that fetches a commit and serves previews and sandboxes; chat and every other surface read the compiled revision only; workspaces with no git remote go. Design new work so it needs no server-side working copy — a change that adds a reason to keep the Factory is going the wrong way. Plan, production data and open decisions: internal-docs/factory-retirement.md.
Read these first when grounding a decision:
internal-docs/multi-instance-fleet.md— how the fleet works today (roles, the stateful-vs-HA matrix, route classification, thesuper_read_onlyguard, graceful degradation, code map). The primary reference.internal-docs/factory-retirement.md— the forward plan: phases, exit tests, and the existing statements it overturns.internal-docs/multi-instance-fleet.md— the original phase ledger + the rejected-alternatives list.internal-docs/compile-boundary.md— operator runbook.
What is built (the current reality)
- Split fleet via
OXY_ROLE(ide | serve | worker | all). One knob selects the topology; everything else derives from it. Aservereplica derives in-process workers + the periodic global driver OFF; every other role ON. The legacy flags (OXY_DISABLE_INPROCESS_WORKERS,OXY_INPROC_GLOBAL_WORKER,--no-workers) survive only as two-directional overrides. - Compile boundary (compile-complete stateless serving). The
idecompiles the working copy → Postgres*_definitionsrows + S3 blobs perrevision_id→ promotes viaworkspaces.current_revision_id. The serve fleet reads the promoted revision and never walks the workspace FS. This is why the read path needs no working copy — and therefore no per-request clone. - Self-routing.
role_manifest::classifyis the single routing authority; a serve replica reverse-proxiesIdeOnlyrequests toOXY_IDE_UPSTREAM(ide_proxy). Replaces the drift-prone external route table that caused three outages. - Route classification + HA carve-out. Every route states its pod at the mount —
route_ide(FS/exec/live-stream) orroute_fleet(Postgres/S3 reads stay HA) — and the two take differently-typed handlers, so a working-copy handler cannot compile onto a fleet route. Plus drift tests and thesuper_read_onlyruntime guard. See theoxy-route-classificationskill. - Runtime-artifact S3 mirror (
runtime_artifact.rs) — charts/results/app-data mirror to the compile-boundary bucket so any replica serves them; ide-down charts degrade to the S3 mirror. - Schedules/monitors fire without a leader. The periodic global driver runs on every eligible node;
tick_schedules/tick_monitor_schedulesCAS-advancenext_run_atso firing is exactly-once across replicas. (Leader election was tried and removed — the CAS already guarantees it, and running on all eligible nodes is better HA.) - Backpressure — admission control (global ceiling + per-tenant fairness, 503 +
Retry-After), worker HPA on outstanding work (queued + claimed) against capacity — see the recipe ininternal-docs/worker-fleet.md; queue depth alone goes absent when the fleet keeps up. - Migrations — a dedicated migrate Job owns the schema;
serve/compile/workerhonourOXY_SKIP_MIGRATIONS; a startup advisory lock serialises co-booting nodes. - Main traffic off the Factory (October 2026). Custom-app function calls run on any replica, with a handler-level replay to the Factory for a workspace with nothing compiled or a database that is a file in the checkout. Airway start, single-window backfill, cancel, live stream and both resets are served by any replica. Custom-app procedure runs execute on the task queue, at most once, with an execution heartbeat.
OXY_IDE_DEFER_QUEUE_WORK(off by default) makes the Factory drive only compiles and take anything else after 30 s unclaimed.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 2d ago First seen · 76 lines · 156 tokens per session scan A 6a13db12c79d
oxy-scaling-design is a skill published in the GitHub repository oxy-hq/oxygen (203 stars, last pushed yesterday), licensed Apache-2.0. It adds 156 tokens to every session and 2,382 once invoked, about $0.0006 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-10-14.
Other skills, from other repositories
gke-compute-classes
Configures, optimizes, and troubleshoots GKE ComputeClasses. Use when configuring Spot VMs with on-demand fallback, targeting specific accelerators (GPUs/TPUs) or machine families, restricting ComputeClass access, or debugging pending pods related to node pool auto-creation. Do not use for cluster-level Node Auto…
gke-reliability
Improves GKE workload reliability, using PDBs, health probes, and topology spread constraints. Use when configuring GKE workload reliability, setting up PDBs, or configuring GKE health probes (liveness, readiness, startup). Don't use for generating K8s YAML manifests (use gke-manifest-generation) or disaster recovery…
gke-workload-security
Audits, configures, and hardens workload-level security controls for Google Kubernetes Engine (GKE) applications and namespaces. Covers running security audits (auditcluster.sh), enforcing Network Policies (default-deny and Dataplane V2 logging), isolating high-risk pods inside GKE Sandbox (gVisor), enforcing Pod…
cloud-architect
Designs cloud architectures, creates migration plans, generates cost optimization recommendations, and produces disaster recovery strategies across AWS, Azure, and GCP. Use when designing cloud architectures, planning migrations, or optimizing multi-cloud deployments. Invoke for Well-Architected Framework, cost…
nemo-automodel-launcher-config
Configure NeMo AutoModel job launches for interactive runs, Slurm clusters, and SkyPilot cloud execution.
azure-mgmt-botservice-dotnet
Azure Resource Manager SDK for Bot Service in .NET. Management plane operations for creating and managing Azure Bot resources, channels (Teams, DirectLine, Slack), and connection settings. Triggers: "Bot Service", "BotResource", "Azure Bot", "DirectLine channel", "Teams channel", "bot management .NET", "create bot".