AgentSkillsCN

observability

引入可诊断的运行时信号（日志/指标/追踪），并附带关联 ID 以及安全的日志记录实践。

SKILL.md
--- frontmatter
name: observability
description: "Add diagnosable runtime signals (logs/metrics/traces) with correlation IDs and safe logging practices."
metadata:
  short-description: Observability plan and checklist

Purpose

Use this skill to make runtime behavior diagnosable by adding logs, metrics, and traces with clear correlation.

When to use

  • •Any change that affects runtime behavior or error handling.
  • •Any time you are unsure which observability signals are required.

How to use

  1. •

    Open references/observability.md and follow the templates.

  2. •

    Define the operations that need to be observable (user-facing or system-facing actions).

  3. •

    Identify correlation identifiers (request_id / job_id / trace_id) and ensure they are logged consistently.

  4. •

    Add the minimum log events: start / outcome / failure, with required fields.

  5. •

    Add metrics for errors and latency (expand to golden signals if relevant).

  6. •

    Add trace spans and ensure logs and metrics are correlated via identifiers.

  7. •

    Apply safety rules (no secrets/PII; follow OWASP/NIST logging guidance).

  8. •

    Control noise (sampling, throttling, or once-only logging).

Output expectation

  • •Record decisions in the Observability Plan.
  • •Ensure the quality gate’s observability checklist passes.