trace-a-request-id
Trace one request identifier across services, queues, retries, and downstream calls to reconstruct its outcome. Use when a distributed operation failed, slowed, duplicated, or became ambiguous.
来源信息
- 仓库
- protosphinx/sphinxstack
- 最近来源活动
- 2026年7月29日 13:55
- 检测到的 SKILL.md 语言
- 英语
- 星标
- 3
- Forks
- 0
安装方式
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
检查来源文件
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
文件资源管理器
2 个文件正在显示 SKILL.md
SKILL.md
来源说明 · 只读预览- name
- trace-a-request-id
- category
- code
- description
- Trace one request identifier across services, queues, retries, and downstream calls to reconstruct its outcome. Use when a distributed operation failed, slowed, duplicated, or became ambiguous.
# trace-a-request-id
Follow one bounded operation end to end without assuming every component propagated context correctly.
## Procedure
1. Confirm the identifier type, source, environment, time, and expected user action.
2. Find the ingress record and capture route, caller-safe context, deployment, and instance.
3. Follow parent, child, correlation, job, and downstream request identifiers.
4. Order spans and logs by synchronized time while accounting for retries and asynchronous handoffs.
5. Mark each boundary's input, output, latency, status, and missing evidence.
6. Separate the original attempt from duplicate, hedged, or retry attempts.
7. Compare the path with one successful request of the same operation.
8. Save a request-path report with the supported failure boundary and next check.
## Guardrails
- Request identifiers are search keys, not authorization credentials.
- Do not infer a complete path when context propagation has gaps.
- Redact associated user, token, payload, and URL data before sharing.
## Done
- A chronological request-path report is saved
- Each service and asynchronous handoff is verified or marked missing
- The final outcome reconciles with user-visible and downstream state
在 GitHub 查看