# Skill Agent Observability

> Define observabilidade para agentes e aplicações de IA com eventos, tracing, custos, latência, qualidade, regressões e privacidade.

- Skill: `iapro-community/skill-agent-observability` (Agent Skill)
- Install (CLI): `npx skillmds@latest add iapro-community/skill-agent-observability`
- Raw SKILL.md: https://api.skillmd.com/api/skills/iapro-community/skill-agent-observability/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: IAPro-Community (https://skillmd.com/u/iapro-community)
- Updated: 2026-09-21
- Page: https://skillmd.com/skills/iapro-community/skill-agent-observability

---


# Agent Observability

Define a minimal event taxonomy for runs, steps, tool calls, retries, failures, tokens, cost, latency, approvals, and outcomes. Redact secrets and personal data. Prefer OpenTelemetry-compatible spans when infrastructure supports it. Add dashboards or reports for success rate, tool failure rate, p50/p95 latency, cost per task, human intervention, and regression comparisons. Do not claim quality improvement without a baseline.