Berserk Skills para Agentes IA
berserkdb/berserk-skills
Skills para debugging, tracing e análise de incidentes via Berserk
Ver no GitHubConjunto de skills e subagentes que capacitam Claude a consultar dados de observabilidade no Berserk. Inclui ferramentas especializadas para análise de logs OpenTelemetry, traces distribuídos, métricas e triagem de incidentes em produção.
Instalação
npx convez-ai berserkdb/berserk-skills1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.
Métricas
Casos de uso
- Investigar erros e exceptions em logs da aplicação
- Analisar traces distribuídos para encontrar gargalos de latência
- Triagem automatizada de incidentes em produção
- Exploração de schema de dados e saúde de serviços
- Análise de métricas OpenTelemetry (gauges, histogramas, somas)
Veredito

Skill robusto para times de observabilidade e SRE que usam Berserk. Automatiza investigações recorrentes de debugging e incident response, economizando tempo em produção. Requer familiaridade com OpenTelemetry e CLI bzrk.
README
A tradução para o português deste README ainda está na fila.
Berserk Skills
Agent skills and subagents for querying Berserk from AI coding agents.
What's Included
| Component | Description |
|---|---|
| berserk skill | Teaches Claude how to use the bzrk CLI — triggered automatically when you mention logs, traces, metrics, or debugging |
| explore agent | General-purpose data exploration — schema discovery, error investigation, service health checks |
| cluster-admin agent | Manage and troubleshoot a Berserk cluster — service health, datasets, segments, ingest tokens, merge tasks |
| otel-log agent | Investigate OpenTelemetry logs — error patterns, log templates, severity analysis, service-level log exploration |
| otel-trace agent | Investigate OpenTelemetry traces — span analysis, latency debugging, trace correlation, service dependency mapping |
| otel-metric agent | Investigate OpenTelemetry metrics — gauge/sum/histogram analysis, metric discovery, time-series queries |
| incident-triage agent | Investigate production incidents — correlate errors, latency spikes, and log patterns to find root cause |
| trace-analysis agent | Analyze distributed traces — build cause-and-effect narratives, identify critical paths and cascading failures |
Claude Code
Install via Plugin Marketplace
/plugin marketplace add berserkdb/berserk-skills
/plugin install berserk@berserk-skills
Install Manually
git clone https://github.com/berserkdb/berserk-skills.git /tmp/berserk-skills
cp -r /tmp/berserk-skills/skills/berserk .claude/skills/
cp -r /tmp/berserk-skills/agents .claude/agents/
Prerequisites
Install the bzrk CLI:
curl -fsSL https://go.bzrk.dev | bash
Configure a profile pointing to your Berserk instance:
bzrk profile list
Usage
Once installed, just ask Claude to query your data:
- "Search for errors in the last hour"
- "Look at traces around 2024-01-07T08:38:00Z"
- "What services are logging?"
- "Investigate connection refused errors"
The agents are invoked automatically when Claude needs to do focused investigation without filling up your main conversation with query results. The berserk skill routes tasks to the appropriate specialist agent based on the query type.
You can also invoke agents explicitly:
- "Use the explore agent to check what services are logging"
- "Use the incident-triage agent to investigate the production outage"
- "Use the trace-analysis agent to analyze trace abc123"
Claude Agent SDK
Use the berserk-explore agent programmatically:
import { query } from "@anthropic-ai/claude-agent-sdk";
for await (const message of query({
prompt: "Investigate error rates across services in the last 6 hours",
options: {
allowedTools: ["Bash", "Read", "Grep", "Glob", "Agent"],
settingSources: ["project"], // loads CLAUDE.md + installed plugins
},
})) {
if ("result" in message) console.log(message.result);
}
Or define agents inline without the plugin:
import { query } from "@anthropic-ai/claude-agent-sdk";
import { readFileSync } from "fs";
const agentPrompt = readFileSync("agents/explore.md", "utf-8");
for await (const message of query({
prompt: "What services have the highest error rates?",
options: {
allowedTools: ["Bash", "Read", "Grep", "Glob", "Agent"],
agents: {
explore: {
description: "Explore and query observability data in Berserk",
prompt: agentPrompt,
tools: ["Bash", "Read", "Grep", "Glob"],
model: "sonnet",
},
},
},
})) {
if ("result" in message) console.log(message.result);
}
OpenCode
OpenCode discovers skills from .claude/skills/ automatically:
git clone https://github.com/berserkdb/berserk-skills.git /tmp/berserk-skills
cp -r /tmp/berserk-skills/skills/berserk .claude/skills/
Or install globally for all projects:
cp -r /tmp/berserk-skills/skills/berserk ~/.claude/skills/
Other Agents
Support for Cursor, GitHub Copilot, and other coding agents is planned. Contributions welcome.