oma-observability
Design or troubleshoot telemetry, SLOs, and incident diagnostics. Route to the relevant signal, system boundary, or vendor guide.
How do I install this agent skill?
npx skills add https://github.com/first-fluke/oh-my-agent --skill oma-observabilityIs this agent skill safe to install?
- Gen Agent Trust Hubpass
The skill is a comprehensive observability assistant designed to help users configure and troubleshoot OpenTelemetry, service meshes, and telemetry signals. It is authored by first-fluke and integrates well with other oma ecosystem skills. The security profile is generally safe; the only notable risk is the surface area for indirect prompt injection, as the skill is designed to ingest and analyze potentially untrusted telemetry data like logs and traces.
- Socketpass
No alerts
- Snykpass
Risk: LOW · No issues
What does this agent skill do?
Observability Agent - Intent-based Router
Scheduling
Goal
Route, design, tune, and review observability work across MELT+P signals, layers, boundaries, vendor categories, transport choices, meta-observability, and incident forensics.
Intent signature
- User asks for observability, telemetry, OTel, metrics, logs, traces, profiles, SLOs, RUM, APM, incident forensics, trace propagation, transport tuning, or observability-as-code.
- User needs vendor/category routing or observability architecture instead of a single vendor's already-covered setup.
When to use
- Setting up an observability pipeline (OTel SDK + Collector + vendor backend)
- Designing traceability across service and domain boundaries (W3C propagators, baggage, multi-tenant, multi-cloud)
- Tuning transport layer (UDP/MTU, OTLP gRPC vs HTTP, Collector DaemonSet vs sidecar topology)
- Running incident forensics (6-dimension localization: code / service / layer / host / region / infra)
- Selecting a vendor category (OSS full-stack vs commercial SaaS vs high-cardinality specialist vs profiling specialist)
- Implementing observability-as-code (Grafana Jsonnet dashboards, PrometheusRule CRD, OpenSLO YAML, SLO burn-rate alerts)
- Meta-observability (pipeline self-health, clock skew detection, cardinality guardrails, retention matrix)
- Covering the MELT+P signal set: metrics, logs, traces, profiles (OTEP 0239), cost (OpenCost), audit (SOC2/ISO), privacy (GDPR/PIPA)
- Evaluating log-pipeline migrations against current upstream support and project requirements
When NOT to use
- LLM ops (prompt versioning, evals, gen_ai span deep dive); use Langfuse, Arize Phoenix, LangSmith, or Braintrust directly
- Data pipeline lineage: use OpenLineage + Marquez, dbt test, or Airflow lineage backends
- IoT / hardware / datacenter physical-layer telemetry (IPMI, BMC, SNMP); use vendor DCIM tooling (Nlyte, Sunbird, Device42)
- Chaos engineering orchestration: use Chaos Mesh, Litmus, Gremlin, or ChaosToolkit (this skill consumes their telemetry; it does not orchestrate chaos)
- GPU / TPU infrastructure observability: use NVIDIA DCGM Exporter + Prometheus
- Software supply chain (SBOM, attestation): use sigstore (cosign / rekor), in-toto framework, SLSA level attestations
- Incident response workflow (on-call rotation, paging, escalation); use PagerDuty, Jira Service Management, or Grafana Cloud IRM
- Full TLS packet inspection: use packet-analysis or vendor TLS inspection tooling
- Security scanning, skill/MCP audits, or penetration testing: use
oma-security(this skill covers security telemetry only: audit trails, WAF signals, PII redaction) - Single-vendor setup already fully covered by that vendor's own published skill; invoke the vendor skill directly
Expected inputs
- Observability intent, target system, architecture boundary, signals, vendor context, and incident symptoms if any
- Existing OTel/collector/vendor configs, dashboards, SLOs, trace/log/metric examples, or deployment topology
Expected outputs
- Routed observability guidance, setup/migration/tuning plan, incident-forensics path, alerting/SLO guidance, or observability-as-code recommendations
- Transport, meta-observability, privacy, audit, and retention checks
- Vendor delegation target when appropriate
Dependencies
Load resources conditionally: start with the execution protocol and the one guide the intent route selects. Read only relevant matrix rows, then add a layer, boundary, signal, or transport guide when its trigger in §References is present. No reference is a mandatory preload.
- OTel/W3C/CNCF references and resources under
resources/ - Vendor categories, matrix, standards, incident forensics, meta-observability, transport, layers, boundaries, and signal guides
- Optional CLI/config tooling from the target stack
Structural Flow
Transitions
- If a vendor-owned skill fully covers setup, delegate instead of duplicating docs.
- If migration is requested, verify upstream support and compatibility before choosing a replacement.
- If incident investigation is requested, use 6-dimensional localization.
- If transport tuning appears, load the one transport guide that matches the observed problem.
Failure and recovery
- If live CNCF/vendor status is load-bearing, verify current status.
- If telemetry samples are missing, provide instrumentation/collection steps before analysis.
- If scope belongs to out-of-scope domains, route to external authoritative tools.
Exit
- Success: observability path is routed, evidence-backed, and checks are explicit.
- Partial success: missing telemetry, stale vendor status, or external-domain handoff is explicit.
Logical Operations
Canonical workflow path
- Classify intent: setup, migrate, investigate, alert, trace, tune, or route. Identify the system boundary the request is about.
- Select layer/boundary/signal coverage from
resources/matrix.md, note coverage gaps, and identify the vendor category when the route needs one. - Load the guide whose trigger in §References matches before producing guidance.
- Read the available evidence: OTel and vendor configs, dashboards, deployment topology, telemetry samples, or incident signals. State which evidence is missing.
- Follow the selected route in §Routes (Intent): route the vendor or category, tune transport, assess meta-observability, or localize the incident. Produce the setup, migration, tuning, alert, trace, or forensics guidance; write observability-as-code, config, or docs only when requested.
- When CNCF/vendor status is load-bearing, verify live state at
https://landscape.cncf.io. - Check pipeline health, clock skew, cardinality, retention, privacy, and audit concerns (Guardrails 3 and 6).
- Report the route, evidence, risks, and handoff references.
Resource scope and effects
Reads observability config, dashboards, alert rules, instrumentation, and the incident symptoms, logs, metrics, traces, and profiles the user supplies. Output is guidance by default; a requested change may modify observability config, dashboards, alert rules, and instrumentation docs. Network access covers vendor or CNCF status checks and telemetry backends when a check needs them. Vendor-specific implementation is routed to vendor-owned skills or external tools.
Guardrails
- Classify intent and use
resources/vendor-categories.mdonly when category selection or vendor delegation is needed. - Load the transport guide matching the observed problem: UDP/MTU, OTLP protocol, Collector topology, or sampling.
- Before declaring a setup complete, verify pipeline health, clock synchronization, cardinality limits, and retention against the project's requirements. Record checks that could not run.
- Verify current upstream support, attribute stability, and migration guidance when they affect the recommendation. Use official project documentation and CNCF status; stored versions are assumptions, not proof of current status.
- Use W3C Trace Context by default, with cloud translations from
resources/boundaries/cross-application.mdwhere required. - Apply PII redaction and sampling-aware baggage controls at collection. Check retention, erasure, and audit requirements for the target system.
- Deliver working configuration or explicitly labeled proposals. Scaffolds and unverified checks cannot count as completed setup.
- Use existing OMA skill names for cross-skill handoffs; do not present future skills as available.
The coverage matrix has four layers (L3, L4, mesh, L7), four boundaries (multi-tenant, cross-application, SLO, release), and seven signals (metrics, logs, traces, profiles, cost, audit, privacy). Read matching rows in resources/matrix.md; N/A cells do not require implementation.
Routes (Intent)
| Intent | Primary target | Fallback |
|---|---|---|
setup | resources/vendor-categories.md → vendor-owned skill | Generic OTel semconv in resources/standards.md |
migrate | Current upstream migration docs + resources/vendor-categories.md log-pipeline section | OTel Collector bridge config |
investigate | resources/incident-forensics.md (MRA + 6-dim localization) | resources/signals/traces.md + resources/signals/logs.md |
alert | resources/boundaries/slo.md (burn-rate alert rules) | resources/observability-as-code.md |
trace | resources/boundaries/cross-application.md (propagator matrix) | resources/layers/mesh.md (zero-code auto-instrumentation) |
tune | The one guide matching the problem: cardinality or series cost → resources/meta-observability.md Section C; telemetry volume → resources/transport/sampling-recipes.md; UDP/MTU, OTLP protocol, or Collector topology → the matching resources/transport/ file | resources/vendor-categories.md when the current vendor's pricing model is the cause |
route | resources/boundaries/multi-tenant.md + resources/transport/collector-topology.md | resources/boundaries/cross-application.md (data residency) |
Record the selected intent route, and any vendor category chosen for setup, with subject observability.route-selection per references/_shared/runtime/route-selection.md.
Invocation and loading
For example: /oma-observability --investigate "5xx spike in checkout".
Other skills pass the intent, system boundary, and evidence; the result returns guidance, checks, and any vendor handoff.
Read resources/execution-protocol.md first. Use the selected route above, add guides by the triggers in §References, then check only the applicable checklist sections before delivery. Do not load all transport, layer, boundary, and signal guides together.
Integrations with OMA Ecosystem
| Skill | Integration point |
|---|---|
oma-debug | Production-failure evidence comes here: pull traces and logs by trace_id (resolve it from request_id when that is all the reporter has) and run resources/incident-forensics.md 6-dimension localization. Reproduction, the fix, and the regression test stay in oma-debug |
oma-qa | Supplies review inputs, not the review: Core Web Vitals targets (INP/LCP/CLS) from resources/layers/L7-application/web-rum.md and backend latency evidence from resources/signals/traces.md. oma-qa owns the performance review and verdict |
oma-tf-infra | Writes and reviews the Terraform that provisions Collectors, backends, and IAM. This skill supplies the requirements: topology (resources/transport/collector-topology.md), transport choice, retention (resources/meta-observability.md Section D), and release metadata. Neither skill ships ready-made observability modules |
oma-security | Source, skill, and MCP scanning and scoped penetration tests. This skill hands over security findings surfaced by telemetry and keeps only the telemetry side: resources/signals/audit.md, resources/layers/L7-application/waf.md, resources/signals/privacy.md |
oma-search | Vendor and upstream documentation lookup (/oma-search --docs "<vendor> <topic>") when no vendor-owned skill is installed, and live CNCF or vendor status checks |
oma-scm | Deployment SHA → service.version OTel attribute + release marker events; see resources/boundaries/release.md |
oma-backend | Propagator and baggage rules (resources/boundaries/cross-application.md); DB N+1 and messaging span patterns in resources/signals/traces.md §2–§3. Back-reference: oma-backend/SKILL.md §References "Observability handoff" |
oma-frontend | resources/layers/L7-application/web-rum.md INP/LCP/CLS checklist. Back-reference: oma-frontend/SKILL.md §References "Observability handoff" |
oma-mobile | resources/layers/L7-application/mobile-rum.md offline-queuing pattern. Back-reference: oma-mobile/SKILL.md §References "Observability handoff" |
oma-db | resources/signals/traces.md §2 DB patterns (N+1, connection pool). Back-reference: oma-db/SKILL.md §References "Observability handoff" |
Versioning & Deprecation
- No per-file version pins:
resources/standards.mdholds the single statement on spec and semconv currency. Where content depends on an attribute's stability tier, the tier is stated inline; confirm it in the semconv registry when it is load-bearing. - Update triggers (not scheduled):
- OTel semconv promotion (Development → RC → Stable) affecting attributes cited in this skill → update
resources/standards.mdand the affected file. - Attribute deprecation → replace the name across all citing files.
- CNCF status change for a vendor/project named in
vendor-categories.md(Graduated / Archived / acquired) → update the vendor table.
- OTel semconv promotion (Development → RC → Stable) affecting attributes cited in this skill → update
- Authoritative live state:
https://landscape.cncf.iofor CNCF project status. This skill does not promise to track it on any schedule; verify at use time if the information is load-bearing.
Before delivering observability changes, check the items of resources/checklist.md that apply, starting with §1 Setup validation; report checks that were not run.
References
Each reference loads on its trigger. Paths are relative to the skill root.
| Trigger | Load |
|---|---|
| Every invocation: flag parsing, intent classification, output shape | resources/execution-protocol.md |
| No flag and the intent is not obvious from the request | resources/intent-rules.md |
| Deciding which artifact belongs at a layer × boundary × signal, or whether a cell is N/A | resources/matrix.md (matching rows only) |
| Attribute naming or stability tier, W3C Trace Context / Baggage rule, OSI scope question, clock synchronization | resources/standards.md |
| Choosing a vendor category or a delegation target | resources/vendor-categories.md |
| Production incident, root-cause localization, missing forensic attributes | resources/incident-forensics.md |
| Collector self-health, clock skew, cardinality budget, retention policy | resources/meta-observability.md |
| Dashboards, alert rules, SLOs, or Collector config managed in git; CI validation of rules | resources/observability-as-code.md |
| Validating a proposed design against known failure patterns | resources/anti-patterns.md (affected category only) |
| Pre-delivery checks | resources/checklist.md (applicable sections only) |
| Unfamiliar output shape for a greenfield setup or a log-agent migration | resources/examples.md |
| StatsD or UDP drops, MTU, fragmentation | resources/transport/udp-statsd-mtu.md |
| OTLP gRPC vs HTTP, proxy or firewall traversal, compression, batch and queue tuning | resources/transport/otlp-grpc-vs-http.md |
| Collector deployment mode, agent vs gateway placement, routing fan-out | resources/transport/collector-topology.md |
| Head or tail sampling policy, reducing trace volume | resources/transport/sampling-recipes.md |
| VPC flow logs, PMTUD black holes, IP-level reachability | resources/layers/L3-network.md |
| TCP retransmits or RTT, load-balancer idle timeouts, QUIC, eBPF instrumentation | resources/layers/L4-transport.md |
| Istio, Linkerd, or Envoy telemetry; zero-code instrumentation; trace breaks at a proxy | resources/layers/mesh.md |
| Browser RUM, Core Web Vitals, source maps, session replay | resources/layers/L7-application/web-rum.md |
| Mobile RUM, offline telemetry queue, app start, ANR | resources/layers/L7-application/mobile-rum.md |
| Crash-free rate, symbolication, crash grouping, crash-to-release correlation | resources/layers/L7-application/crash-analytics.md |
| Edge 403 storm, WAF false positives, ruleset rollout, WAF fail-open or fail-closed | resources/layers/L7-application/waf.md |
| Tenant isolation, per-tenant routing or sampling, data residency | resources/boundaries/multi-tenant.md |
Propagators, baggage, cross-cloud or cross-vendor trace continuity, request_id, third-party dependency outage | resources/boundaries/cross-application.md |
| SLI or SLO definition, error budget, burn-rate alerts | resources/boundaries/slo.md |
| Release markers, canary analysis, feature-flag telemetry, GitOps drift | resources/boundaries/release.md |
| Metric naming, histogram buckets, exemplars, label design | resources/signals/metrics.md |
| Log agent or pipeline choice, structured logging, trace-ID injection into logs | resources/signals/logs.md |
| Span design; DB (N+1, connection pool) and messaging span patterns | resources/signals/traces.md |
| Continuous profiling, flame graphs, heap or CPU diff across releases | resources/signals/profiles.md |
| Cost attribution, OpenCost, FinOps unit economics | resources/signals/cost.md |
| Audit trail, compliance evidence, retention schedule, tamper evidence | resources/signals/audit.md |
| PII redaction, pseudonymization, cross-border transfer, backend access control | resources/signals/privacy.md |
How can the creator link this skill?
Add the canonical catalog link to the repository README so users can inspect current installs and available audits. The publishing guide covers the complete discovery path.
<a href="https://skillzs.dev/skills/first-fluke/oh-my-agent/oma-observability">View oma-observability on skillZs</a>