Skip to main content

plan-an-incident-exercise

Design a realistic, safe, and measurable incident exercise. Use when a team needs to test detection, decisions, escalation, containment, recovery, communication, or evidence without creating uncontrolled production risk.

ソース情報

リポジトリ
protosphinx/sphinxstack
ソースの最終更新活動
2026年7月29日 13:55
検出された SKILL.md の言語
英語
スター
3
フォーク
0

インストール方法

デフォルトでは、最初にソースを確認する Prompt が選択されています。直接コマンドに切り替えるか、ローカルコピーをダウンロードすることもできます。

ソースファイルを確認

インストールを決める前に、SKILL.md と SkillsMP に表示されている付属ファイルをお読みください。

ファイルエクスプローラー
2 ファイル

SKILL.md を表示中

SKILL.md
ソースの指示 · 読み取り専用プレビュー
name
plan-an-incident-exercise
category
start
description
Design a realistic, safe, and measurable incident exercise. Use when a team needs to test detection, decisions, escalation, containment, recovery, communication, or evidence without creating uncontrolled production risk.
# plan-an-incident-exercise Create a scenario that discriminates capability from confidence. Test a few important objectives deeply. ## Procedure 1. Choose sponsor, exercise director, safety controller, participants, observers, and independent evaluator. 2. Define two to five measurable objectives and the evidence that would demonstrate each one. 3. Select a scenario tied to real services, dependencies, and known risk without reproducing active harm. 4. Set starting state, timeline, injects, expected decisions, branch conditions, and information available to each role. 5. Mark every artifact and channel as exercise-only and define the real-incident interruption protocol. 6. Establish production, privacy, personnel, customer, vendor, and physical-safety boundaries. 7. Prepare observation forms for decision time, authority, evidence, handoffs, tools, communications, and unmet assumptions. 8. Rehearse facilitation, inject delivery, technical simulation, stop conditions, and cleanup. 9. Brief participants on safety and purpose without revealing the expected path. 10. Define debrief, finding severity, action ownership, acceptance tests, and targeted retest timing. ## Guardrails - Never introduce uncontrolled load, malware, data loss, payment, customer contact, or emergency-service activity. - Do not design the exercise to embarrass one person or reward memorized answers. - Keep scenario data synthetic or approved and minimized. - Pause when participants cannot distinguish exercise from reality. ## Done - Objectives map to observable evidence - Scenario and injects test realistic decisions - Safety, stop, real-incident, and cleanup controls are rehearsed - Evaluation and remediation paths exist before execution
GitHubで見る