OBSERVABILITY / Elastic
LLM and Agentic Observability
Answers LLM latency, token, cost and call-chain questions from telemetry already in Elastic, and says so when cost is not instrumented.
“Answer questions about LLM and agentic-application behavior from data already ingested into Elastic: latency and error rate, token and cost utilization, response quality and guardrail events, and agentic call-chain orchestration. Use when the user asks about LLM monitoring, GenAI observability, token spend or AI cost, model latency, prompt or guardrail failures, or how an agent's tool-call chain executed.”
01 / THE REASONING
Why this made the selection.
- If no cost field exists, the file says to report that cost is not instrumented and never multiply token counts by a guessed per-token price.
- Zero rows means data is missing or not collected, never that the condition is healthy, and ingestion paths and GenAI field names are discovered rather than assumed.
02 / THE REVIEW RECORD
What we actually inspected.
Source review has boundaries.
A clear record is more useful than a “safe” badge.
Material inspected
- skills/observability/llm-obs/SKILL.md (compatibility, environment, output discipline, data locations, discovery process, cost questions, operations table)
- README.md (skill table, security considerations, getting started)
- LICENSE Apache-2.0, copyright 2026 Elasticsearch BV
Our findings
- Only traces can rebuild a call chain; with integration metrics alone the file says to state that the chain cannot be reconstructed.
- The file notes that cost is not part of the OpenTelemetry GenAI specification.
Not established by this review
- The installer was not run, and the skill was not executed against a project or service. No query was run and no deployment was contacted.
- references/esql-recipes.md, genai-fields.md and integrations.md were not opened.
- Not executed: it was not installed or run.
The review applies to the material and revision named here. A newer upstream release can change its behavior.
03 / PUT IT TO WORK
Add a skill to your workflow.
DOCUMENTED COMMAND
claude plugin marketplace add https://github.com/elastic/agent-skills Copying does not execute this command. It may retrieve a newer version than the reviewed source.
- The README then installs the observability group with claude plugin install elastic-observability@elastic-agent-skills, and advises installing only the skills you need.
- Neither command pins revision baa511126ba2dc37b52e273b52734f8e4e0d323c. Compare the installed SKILL.md and references with that revision.
- The README advises minimum-privilege API keys with read-only access as the default.
Before you start
- Claude Code for the plugin path
- The elastic CLI (>= 0.2) configured for a deployment holding LLM telemetry
04 / FOLLOW THE EVIDENCE
The source trail.
Our notes are separate from the original resource.
Check upstream before adopting a new version.
- Skill definition ↗Checked
https://github.com/elastic/agent-skills/blob/baa511126ba2dc37b52e273b52734f8e4e0d323c/skills/observability/llm-obs/SKILL.md
Supports: summary, upstreamDescription, whySelected, limitations, access, review
https://github.com/elastic/agent-skills/blob/baa511126ba2dc37b52e273b52734f8e4e0d323c/README.md
Supports: install, compatibility
- Repository licence ↗Checked
https://github.com/elastic/agent-skills/blob/baa511126ba2dc37b52e273b52734f8e4e0d323c/LICENSE
Supports: license
https://raw.githubusercontent.com/elastic/agent-skills/baa511126ba2dc37b52e273b52734f8e4e0d323c/skills/observability/llm-obs/SKILL.md
Supports: upstreamDescription, review.notTested, limitations
https://raw.githubusercontent.com/elastic/agent-skills/baa511126ba2dc37b52e273b52734f8e4e0d323c/README.md
Supports: install.command