- Home
- Skills
- DevOps & Cloud
- Incident Postmortem Writer: Blameless Report from Your Logs
More screenshots
Works with the AI Agents you already use
Incident Postmortem Writer: Blameless Report from Your Logs
The agent drafts a blameless postmortem for your team to review before publishing internally, a short update for customers, and the action items as a CSV.
$4
Incident Postmortem Writer: Blameless Report from Your Logs
Example session with this skill installed
Write a blameless postmortem for Saturday's checkout outage from the files in incident/: the chat export is office time (UTC+3), alerts, app log and deploy history are UTC. Add something short for the status page. Owners can be the orders team, platform team, on-call lead or support lead.
- Read your context and instructions
- Compiled the incident postmortem writer
Root cause and contributing factors
Trigger: release 412 of orders-api, deployed 10:36:02-10:41:07 (deploys L4-L5).
Causal chain
- Release 412 updated the database driver from 8.11 to 8.13 and included a "pool config cleanup"
(deploys L4), so the pool configuration of the service changed. (evidence) - Instances on release 412 created pools of 5 connections; instances on release 411 create pools of 20
(app L3, L11, L14 against L1, L18), so each instance could serve a quarter of the concurrent database work.
(evidence) - The reason for the value 5: the explicit setting of 20 was removed and the new driver's default is 5.
(inferred from the release author's recollection, chat L15; the diff was not available) - Saturday midday demand needed more than 5 connections per instance, so requests queued (up to 41 waiters
on one instance, app L8), waited the 5000 ms acquire timeout and returned HTTP 503. (evidence) - The pipeline's unit, integration and smoke checks passed (deploys L5), so nothing stopped the release. The
smoke test sends one request at a time (notes L12, reported), which a pool of 5 serves without queueing.
(evidence for the pass; reported for the reason)
Root cause (most consistent with the evidence; the release 412 diff would confirm link 3): the effective
connection pool size of orders-api is not fixed or checked anywhere. It can change through a library default,
and no step between merge and full production traffic compares the effective value with the previous release
or exercises the service under concurrent load.
Connects securely to your tools. The creator never sees your data.
About this skill
English
For engineering teams who have to explain an incident to people who were not there.
Give your agent the raw evidence: the incident channel export, alerts, logs, deploy history and notes. The agent drafts a blameless postmortem for your team to review before publishing internally, a short update for customers, and the action items as a CSV.
What it writes each run
postmortem.md: a plain summary, impact with numbers taken only from the evidence, a timeline in one time zone with a source for every row, detection and response, a numbered causal chain that separates trigger, root cause and contributing factors, what went well, what went badly, where you got lucky, action items, open questions and a gap list.status_update.md: a short customer-facing update, drafted without internal names or blame for you to review.action_items.csv: the same action items with id, type, owner role, priority and done-condition.
Timeline first, from the evidence
A bundled Python script (standard library, no network calls) reads timestamped lines from several files, including ISO 8601, web server and syslog formats, epoch times and time-only chat lines. It converts them to one time zone from the offsets you give, sorts them, and flags long silences, out-of-order lines, minute-precision times and assumed zones. The agent builds the timeline before any cause is discussed.
Facts apart from interpretation
The skill instructs the agent to keep summary, impact, timeline and detection to what the evidence shows, to cite evidence for each link of the causal chain or mark it inferred, to show both sides when sources disagree, and to write an unsupported cause as not established.
Blameless by rule
The skill instructs the agent to refer to roles and systems in place of named people, to rewrite phrases such as "should have" and "failed to" from a bundled table, and to keep "human error" out of the chain. Review the draft for names and blame wording before sharing.
Honest about its limits
It works from what you provide, and the agent is instructed to mark gaps [NEEDS: ...]. Line references are written by the agent and should be spot-checked by someone who was there. The customer update is a draft, and formal regulatory or contractual notices are outside its scope.
Arabic
On request all three files are written in Arabic; system names, commands, log excerpts and the CSV header stay in Latin script. Arabic output was tried on one test case and was not reviewed by a second native reader; check it before sharing.
In the download: SKILL.md, the script, four reference files (method, blameless language, action items, Arabic notes), three templates, and one complete example: five evidence files from a database pool incident in, the three deliverables out, with nine action items and the raw merged timeline.
How it was tried: Tried on two cases (a database pool incident in English, a certificate expiry incident in Arabic); the download includes one worked example in English. The bundled script has 17 unit tests.
بالعربي
لفرق الهندسة التي عليها أن تشرح حادثة لمن لم يحضرها.
أعطِ وكيلك الأدلة الخام: تصدير قناة الحادثة، والتنبيهات، والسجلات، وسجل الإصدارات، والملاحظات. تستلم تقرير ما بعد الحادثة بلا لوم، ونصًّا قصيرًا للعملاء، وقائمة الإجراءات في ملف CSV.
ماذا تستلم
- ملف
postmortem.md: ملخص، وأثر بأرقام من الأدلة فقط، وتسلسل زمني بتوقيت واحد مع مصدر لكل سطر، وسلسلة سببية تفصل الشرارة عن السبب الجذري والعوامل المساعدة، وأسئلة مفتوحة. - ملف
status_update.md: تحديث موجز للعملاء بلا أسماء داخلية. - ملف
action_items.csv: الإجراءات، ولكل إجراء نوع ودور مسؤول وأولوية وشرط إنجاز يمكن التحقق منه.
التسلسل الزمني أولًا
سكربت بايثون مرفق يعمل على جهازك بلا اتصال بالشبكة: يقرأ الأسطر ذات الأوقات من عدة ملفات، ويحوّلها إلى توقيت واحد، ويرتّبها، ويعلّم الفجوات والأسطر غير المرتبة.
بلا لوم وبلا تخمين
تنص التعليمات على الإشارة إلى الأدوار لا إلى الأشخاص، وعلى ألا يُذكر سبب لا تسنده الأدلة بل يُكتب أنه غير مثبت. راجع المسودة قبل مشاركتها.
حدود واضحة
التقرير مسودة للمراجعة، وأرقام الأسطر المذكورة تحتاج تدقيقًا ممن حضر الحادثة. وتحديث العملاء مسودة، وقرار النشر والتعويض لك.
العربية
عند الطلب تُكتب الملفات الثلاثة بالعربية، وتبقى أسماء الأنظمة والأوامر ومقتطفات السجلات بالحروف اللاتينية.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for DevOps & Cloud.
Works only from the evidence you provide: where it does not show a cause, a number or a time, the skill instructs the agent to say so and mark it [NEEDS: ...], and to write a cause the evidence does not establish as not established. Line references in the timeline are written by the agent and nothing checks them mechanically, so have someone who was there spot-check them. The script does not read ambiguous numeric dates such as 03/10/2026, needs the year for syslog-style lines, and places minute-precision chat lines before second-precision lines of the same minute (flagged). Both test incidents were small (under 70 timed lines each), so behaviour on very long logs is not shown by a worked case. Action items carry owner roles and priorities, not names or due dates, and whether a done-condition fits your environment is your call. The customer update is a draft; compensation, contractual notices and formal regulatory or security reports stay with you. Arabic output was read only by the agent that wrote it, not by a second native reader. التقرير مسودة للمراجعة مبنية على أدلتك فقط.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 3 days ago
- Software
python3 (3.10 or newer)
Permissions
Reads the evidence files you point it at (chat export, alerts, logs, deploy history, notes) and does not change them. Writes postmortem.md, status_update.md, action_items.csv and a work/timeline_raw.md working file into a folder in your project. Runs its bundled Python script locally (timeline_tools.py); it uses the standard library only and makes no network calls. The working timeline contains names and raw chat text and is not meant to be published. يقرأ ملفات الأدلة ويكتب ملفات التقرير فقط، بلا اتصال بالشبكة.