NVIDIA/NeMo-Relay is a runtime and library that gives coding agents and applications shared control over agent execution scopes, policies, plugins, lifecycle events, and observability data. It is for developers who need to inspect or instrument runs from agents such as Codex or Claude Code, or integrate frameworks and export traces and trajectories. The catalogue entries provide Relay plugins, hooks, instructions, skills, and an MCP server for those workflows.
Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add NVIDIA/NeMo-Relay --skill nemo-relay-instrument-context-isolationgit clone --depth 1 https://github.com/NVIDIA/NeMo-RelayWrote 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/nvidia/nemo-relay/nemo-relay-instrument-context-isolation)<a href="https://agentmods.dev/skills/nvidia/nemo-relay/nemo-relay-instrument-context-isolation"><img src="https://agentmods.dev/badge/skills/nvidia/nemo-relay/nemo-relay-instrument-context-isolation/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/nvidia/nemo-relay/nemo-relay-instrument-context-isolation"><img src="https://agentmods.dev/badge/skills/nvidia/nemo-relay/nemo-relay-instrument-context-isolation.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
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 | $0.00041 | $0.00585 |
| Opus 5 | $0.00020 | $0.00293 |
| Sonnet 5 | $0.00008 | $0.00117 |
| Haiku 4.5 | $0.00004 | $0.00059 |
Grade A, and why
nemo-relay-instrument-context-isolation 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 10d 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 — 64 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Use Context Isolation
Use this skill when an application runs concurrent requests, worker pools, async tasks, goroutines, or multiple agents in the same process. Treat scope-stack ownership as part of the request boundary.
Core Rule
Each independent request, agent, or workflow needs its own scope stack. Do not share one mutable stack across unrelated concurrent work unless you want shared ancestry and shared scope-local middleware.
Embedded Scope Model
- A root scope is always present, and pushed scopes form a parent-child tree beneath it.
- Scope hierarchy determines event parentage, lifetime boundaries, and visibility for scope-local middleware and subscribers.
- Standard scope types include
Agent,Function,Tool,Llm,Retriever,Embedder,Reranker,Guardrail,Evaluator,Custom, andUnknown. - Scope start and end events can carry semantic
inputandoutputpayloads when the scope represents a request, task, or meaningful result boundary. - Scope-local registrations disappear when the owning scope closes; use them for behavior that should not outlive a request or agent run.
- Mark events are useful for retries, checkpoints, interrupts, and important state transitions that are not full spans.
Per-Language Defaults
- Python: rely on task-local behavior via
get_scope_stack()andcontextvars, or explicitly propagate when work leaves the current execution context - Rust core: use runtime helpers such as
create_scope_stack(),current_scope_stack(), andset_thread_scope_stack(...)when an integration needs explicit stack ownership - Go: use
NewScopeStack()andScopeStack.Run(...)for goroutine-safe usage - Node.js: create and set a scope stack explicitly for the current execution
path with
createScopeStack()andsetThreadScopeStack(...)
Common Failures
- Events from different requests appear under one root UUID
- Scope-local middleware leaks across requests
- Worker-thread work runs without the expected active scope
- Integrations activate NeMo Relay without an explicitly initialized stack
- Relying on a thread-local stack after crossing async tasks, goroutines, or JS worker boundaries
What ships with it
4 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- 10d ago First seen · 64 lines · 41 tokens per session scan A 3a6faa4fc1f5
nemo-relay-instrument-context-isolation is a skill published in the GitHub repository NVIDIA/NeMo-Relay (167 stars, last pushed yesterday), licensed Apache-2.0. It adds 41 tokens to every session and 585 once invoked, about $0.0002 per session on Opus 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 2026-08-30.
Other skills, from other repositories
cqrs-implementation
Implement Command Query Responsibility Segregation for scalable architectures. Use when separating read and write models, optimizing query performance, or building event-sourced systems.
workflow-orchestration-patterns
Design durable workflows with Temporal for distributed systems. Covers workflow vs activity separation, saga patterns, state management, and determinism constraints. Use when building long-running processes, distributed transactions, or microservice orchestration.
microservices-patterns
Design microservices architectures with service boundaries, event-driven communication, and resilience patterns. Use when building distributed systems, decomposing monoliths, or implementing microservices.
python-sdk
Implement or modify Python SDK behavior under python/composio, including tools, toolkits, sessions, auth configs, connected accounts, client integration, and shared Python models. Use for Python core runtime/API work; pair with python-testing and cross-sdk-parity when TypeScript must match.
azure-eventhub-dotnet
Azure Event Hubs SDK for .NET. Use for high-throughput event streaming: sending events (EventHubProducerClient, EventHubBufferedProducerClient), receiving events (EventProcessorClient with checkpointing), partition management, and real-time data ingestion. Triggers: "Event Hubs", "event streaming"…
azure-search-documents-dotnet
Azure AI Search SDK for .NET (Azure.Search.Documents). Use for building search applications with full-text, vector, semantic, and hybrid search. Covers SearchClient (queries, document CRUD), SearchIndexClient (index management), and SearchIndexerClient (indexers, skillsets). Triggers: "Azure Search .NET"…