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/etr/groundwork/instrument-observabilitynpx skills add etr/groundwork --skill instrument-observabilitygit clone --depth 1 https://github.com/etr/groundworkWhat 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 | $0.00038 | $0.00806 |
| Opus 5 | $0.00019 | $0.00403 |
| Sonnet 5 | $0.00008 | $0.00161 |
| Haiku 4.5 | $0.00004 | $0.00081 |
Grade A, and why
instrument-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 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 — 60 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Instrument Observability
Overview
Observability is added with the code that needs it, not bolted on after an outage. Shift Left: the cheapest time to make a change observable is while you still hold its context — what can fail, what "normal" looks like, which boundary the latency lives behind.
Core principle: A change you cannot observe in production is a change you cannot operate. Logs, metrics, traces, and alerts are part of "done," not a follow-up ticket.
When to Use
- Adding or changing a service boundary, request handler, job, or external call
- Code paths that can fail partially, retry, time out, or degrade
- Any change whose health you would want to confirm during a [[staged-rollout]]
Skip only for changes with no runtime behavior (docs, pure refactors with identical I/O, config-only edits).
Process
Apply each layer to the change.
- Structured logging. Emit machine-parseable events (key-value / JSON), not interpolated prose. Attach the correlation/trace id and the dimensions you'd filter on (tenant, route, outcome). Log decisions and failures, not control flow.
- RED metrics. For every request-serving surface, instrument the three:
- Rate — requests per second handled
- Errors — failed requests per second (and the error class)
- Duration — latency distribution (histogram, so you get p50/p95/p99 — never a single mean)
- Trace spans. Wrap each external boundary (DB query, RPC, queue, third-party API) in a span that propagates context. Spans turn "the request was slow" into "the request was slow here."
- Symptom-based alerts. Alert on user-visible symptoms (error rate breached, latency SLO burning), not on causes (CPU high, pod restarted). Causes generate noise; symptoms generate pages worth waking for. Each alert names the symptom and points at the dashboard/runbook.
Rationalizations
| Excuse | Reality |
|---|---|
| "I'll add metrics once it's in prod" | The first incident is the worst time to discover you're blind. Shift Left. |
| "The mean latency is fine" | A mean hides the p99 tail where users actually hurt. Use a histogram. |
| "There are already logs" | Unstructured logs you can't query are not observability. Structure them. |
| "I'll alert on CPU and disk" | Cause-based alerts page you for non-problems and miss real ones. Alert on symptoms. |
| "Tracing is a separate project" | One span around each boundary is minutes of work and the only thing that localizes latency. |
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 · 60 lines · 38 tokens per session scan A 4c5672848430
instrument-observability is a skill published in the GitHub repository etr/groundwork (42 stars, last pushed 21d ago), licensed MIT. It adds 38 tokens to every session and 806 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
kagent-dev
Development guide for kagent's v1alpha3 Harness and AgentTemplate CRDs, AgentInstance gRPC control plane, upstream A2A integration, Substrate runtime provisioning, tests, generation, and PR workflow. Use for any implementation, debugging, review, or CI task in the kagent repository.
timoni
Use when deploying applications to Kubernetes with Timoni. Covers installing and upgrading module instances from OCI registries, composing multi-app deployments with bundles, injecting values from clusters or CI with runtimes, targeting multiple clusters, and authoring, testing, signing and publishing modules with CUE.
azmon-mirroredcatalogs-operations-cli
Brings Azure Monitor, Application Insights, and Log Analytics telemetry into Fabric as Eventhouse external delta tables and correlates it with business data. Use to onboard observability data, judge whether latency or availability affected revenue, or build a Real-Time dashboard and Operations Agent over it.
durable-objects
Create and review Cloudflare Durable Objects. Use when building stateful coordination (chat rooms, multiplayer games, booking systems), implementing RPC methods, SQLite storage, alarms, WebSockets, or reviewing DO code for best practices. Covers Workers integration, wrangler config, and testing with Vitest. Biases…
ship-web-games
Package, deploy, and verify a playable Three.js or web game. Use for release builds, asset delivery, private/public deployment, production smoke tests, browser proof, release notes, rollback readiness, and cleanup of temporary QA resources.
hyperpod-version-checker
Check and compare software component versions on SageMaker HyperPod cluster nodes - NVIDIA drivers, CUDA toolkit, cuDNN, NCCL, EFA, AWS OFI NCCL, GDRCopy, MPI, Neuron SDK (Trainium/Inferentia), Python, and PyTorch. Use when checking component versions, verifying CUDA/driver compatibility, detecting version mismatches…