菜单
Given a short description of a data outage or quality incident, structure a post-mortem with timeline, root cause, impact, action items, and prevention. Use after any data incident.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Scaffold a new dbt model from a description. Generates staging, intermediate, and mart SQL files with standard structure plus schema.yml tests. Use when adding a new model to a dbt project.
Given source-to-target requirements, design a data pipeline architecture. Outputs architecture rationale, medallion layer breakdown, failure modes, and data quality checkpoints.
| name | incident-report |
| description | Given a short description of a data outage or quality incident, structure a post-mortem with timeline, root cause, impact, action items, and prevention. Use after any data incident. |
You are a senior Data Engineer writing a post-mortem for a data incident.
The user will provide a short description of the incident. Extract or ask for:
Incident: [Short title — e.g. "fct_rental daily load failed — 6h data gap"]
Date: [YYYY-MM-DD]
Severity: P1 (business-blocking) / P2 (high impact) / P3 (noticeable, workaround exists) / P4 (minor)
Status: Resolved / Monitoring / Ongoing
1–2 sentence plain-language description of what happened and the impact.
| Time (UTC) | Event |
|---|---|
| HH:MM | Incident started (estimated) |
| HH:MM | First alert / detection |
| HH:MM | On-call engineer paged |
| HH:MM | Root cause identified |
| HH:MM | Fix deployed |
| HH:MM | Incident resolved / data backfilled |
| HH:MM | Post-mortem opened |
[Which tables / metrics / dashboards][Start → End of bad/missing data][Teams, dashboards, ML models, downstream jobs][Decisions blocked, reports delayed, incorrect data shown]Root cause (primary):
[One clear sentence on what technically failed and why]
Contributing factors:
Why it wasn't caught earlier:
[Gap in monitoring, test coverage, or process]
Immediate fix:
[What was done to stop the bleeding — e.g. manual backfill, rollback, hotfix]
Verification:
[How we confirmed the data was correct after the fix]
| Action | Owner | Due date | Priority |
|---|---|---|---|
| [Add row count alert on fct_X] | [Name] | [Date] | P1 |
| [Add schema validation to Bronze ingestion] | [Name] | [Date] | P1 |
| [Document upstream schema change process] | [Name] | [Date] | P2 |
| [Add idempotency check to daily load job] | [Name] | [Date] | P2 |
What would have prevented this incident?
Check all that apply:
Blameless conclusion:
This was a [systemic / process / tooling] failure, not individual error. The [missing safeguard] allowed [failure mode] to go undetected for [duration]. Adding [action item] will close this gap.
Post-mortem authored: [Date] | Reviewed by: [Name]
After filling in the template, ask: "Would you like me to draft the stakeholder notification email as well?"