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 agentmods add skills/nvidia/nemo-relay/maintain-observabilitynpx skills add NVIDIA/NeMo-Relay --skill maintain-observabilitygit 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/maintain-observability)<a href="https://agentmods.dev/skills/nvidia/nemo-relay/maintain-observability"><img src="https://agentmods.dev/badge/skills/nvidia/nemo-relay/maintain-observability.svg" alt="Measured on agentmods" 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 | $0.00023 | $0.00789 |
| Opus 5 | $0.00012 | $0.00394 |
| Sonnet 5 | $0.00005 | $0.00158 |
| Haiku 4.5 | $0.00002 | $0.00079 |
Grade A, and why
maintain-observability 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 4d 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 — 79 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Maintain Observability Surfaces
Companion Guidance
Use karpathy-guidelines alongside this skill for implementation or review
work. Keep changes scoped, surface assumptions, and define focused validation
before editing.
Use this skill when changing event fields, exporter behavior, subscriber config,
or binding parity for ATIF or the full, gen_ai, and openinference
OpenTelemetry projections.
Surfaces To Keep In Sync
- Core event model and emitted fields
crates/core/src/observability/atif.rscrates/core/src/observability/otel.rscrates/core/src/observability/openinference.rs- FFI and binding-native wrappers where the config or lifecycle is exposed
- Python, Go, and Node.js config objects and subscriber/exporter methods
- Observability config version 3, where one
opentelemetrysection contains typed endpoints and OpenInference has no standalone public surface - Docs under
docs/about-nemo-relay/concepts/subscribers.mdxanddocs/configure-plugins/observability/
Design Checklist
- Is this an event-model change, exporter-config change, or lifecycle change?
- Do all bindings expose the same logical knobs and semantics?
- Does every OpenTelemetry endpoint require a type and nonblank destination?
- Does each endpoint resolve
header_envvalues at activation and reject missing, blank, or duplicate headers? - Do layered ATOF sink, ATIF storage, and OpenTelemetry endpoint lists concatenate with higher-precedence entries first?
- Are OpenTelemetry and OpenInference dependencies unconditional rather than Cargo feature-gated?
- Does
gen_aiavoidnemo_relay.*, project sanitized LLM instructions and messages into the standard content attributes, and emit minimal spans for scopes without GenAI semantics so their parentage is preserved? - Does
enable_full_payloadspreserve complete sanitized LLM request input and annotations while leaving credential removal and sanitizers active? - Does Relay derive compliant trace and span IDs consistently across typed OpenTelemetry endpoints while preserving lifecycle parentage?
- Are mark events, start/end events, and orphan cases still handled correctly?
- Does a sanitized tool result annotation remain opaque under
category_profile.tool_result_annotation, ATIF observation-resultextra.tool_result_annotation, and the singlenemo_relay.tool.result.annotationattribute infullandopeninference, whilegen_aiomits it? - Do examples and docs use each exporter's documented flush/deregister order before shutdown?
- Are span or trajectory fields still derived from the intended event data?
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.
- 4d ago First seen · 79 lines · 23 tokens per session scan A f9adfb3d613e
maintain-observability is a skill published in the GitHub repository NVIDIA/NeMo-Relay (140 stars, last pushed today), licensed Apache-2.0. It adds 23 tokens to every session and 789 once invoked, about $0.0001 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
atifact
This skill should be used when the user asks to "extract agent trajectory", "convert HAR to trajectory", "get trajectory from session", "parse agent session", "convert Claude Code logs", "convert Copilot CLI logs", "extract ATIF", or needs to convert HAR files or JSONL session logs into ATIF trajectory JSON using the…
bump-version
This skill should be used when the user asks to "bump the version", "release a new version", "prepare a release", "update the version number", "version bump", or needs to determine and apply a semver version increment based on recent commits.
systematic-debugging
Use when encountering any bug, test failure, or unexpected behavior, before proposing fixes.
brainstorming
You MUST use this before any creative work - creating features, building components, adding functionality, or modifying behavior. Explores user intent, requirements and design before implementation.
auto-perf-optimize
Run agent-driven VS Code performance or memory investigations. Use when asked to launch Code OSS, automate a VS Code scenario, run the Chat memory smoke runner, capture renderer heap snapshots, take workflow screenshots, compare run summaries, or drive a repeatable scenario before heap-snapshot analysis.
chat-perf
Run chat perf benchmarks and memory leak checks against the local dev build or any published VS Code version. Use when investigating chat rendering regressions, validating perf-sensitive changes to chat UI, or checking for memory leaks in the chat response pipeline.