incident · datawithbaraa — Claude Code skill: job failure analysis
The incident Claude Code skill traces a failed job run from its error toward a root cause. Give it a failure date or job and run; it produces an HTML page with a workaround, solution options and email drafts.
MIT ★ 106 updated 2026-11-06
/plugin marketplace add DataWithBaraa/claude-code-data-engineering-skills
/plugin install data-engineering@data-with-baraaWhat incident does
The skill gathers the failed task, file and line, error message, impact on consumers, and check results for each layer of the run. It reads run records, quarantine data, delivered files, pipeline code and the incident log to look for similar failures and changes that may explain the error.
The result is a single HTML file with a section for each part of the investigation. It includes two solution options with effort estimates and a recommendation, future risks across tables, and draft emails to consumers and the source owner. The skill does not change code, rerun jobs or send emails; it adds an incident-log entry only after approval.
What it helps with
- locate the task and error behind a failed job run
- compare a failure with earlier incidents
- describe a workaround for getting fresh data
- weigh two fixes and estimate their effort
- draft incident emails and a log entry
Example prompts for the agent
- Analyze the failed job run from 2025-02-14 and create an incident page.
- Find the cause of the error in the
daily_ordersrun and propose two fixes. - Draft emails to the consumers and source owner about this incident.
How to install incident
1Plugin marketplace
/plugin marketplace add DataWithBaraa/claude-code-data-engineering-skills/plugin install data-engineering@data-with-baraaThis installs the whole data-engineering plugin with all its skills
2Skills CLI (Vercel)
npx skills add DataWithBaraa/claude-code-data-engineering-skills --skill incident -a claude-code3Manually: put the files in place
~/.claude/skills/incident/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/incident/
1Skills CLI (Vercel)
npx skills add DataWithBaraa/claude-code-data-engineering-skills --skill incident -a codex2Manually: put the files in place
.agents/skills/incident/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add DataWithBaraa/claude-code-data-engineering-skills --skill incident -a cursor2Manually: put the files in place
.cursor/skills/incident/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install DataWithBaraa/claude-code-data-engineering-skills skills/incident2Manually: put the files in place
.github/skills/incident/SKILL.md1Skills CLI (Vercel)
npx skills add DataWithBaraa/claude-code-data-engineering-skills --skill incident -a gemini-cli2Platform CLI
gemini skills install https://github.com/DataWithBaraa/claude-code-data-engineering-skills --path skills/incidentManually: put the files in place
.devin/skills/incident/SKILL.mdThe legacy .windsurf/skills/ path still works
Claude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about incident
How do I install incident in Claude Code?
Run /plugin marketplace add DataWithBaraa/claude-code-data-engineering-skills, then /plugin install data-engineering@data-with-baraa.
What does incident do?
It investigates a failed job run and creates an HTML report covering the likely cause, a workaround, two proposed fixes, future risks and draft emails.
Can incident fix the job or send emails?
No. It does not edit code, rerun jobs or send emails. It writes an incident-log entry only after approval.
Source text
SKILL.md · MIT
--- name: incident description: Use when a job run failed with an error, to walk the incident from the error to the root cause, the workaround, two solutions, the future risks, the email drafts and the log entry. argument-hint: "[date or run]" --- # Incident The job: from the first error to the log entry. It proposes and drafts. It never fixes and never sends. Input: the date of the failed run, or the job and its run. ## Steps 1. Where it failed: the job, the task, the file and line, and the error message. 2. Status: the severity, the impact on the consumers, and the next step. 3. What happened: the run layer by layer, with each check and its result. 4. Root cause: what changed, with the rows before and after. 5. Seen before: check the incident log for the same kind of failure and its known fix. 6. Workaround: how the consumers get fresh data today. 7. Solutions: 2 options with the effort of each, and which one to recommend. 8. Future risks: every table with the same weakness, and which ones are safe. 9. Email drafts: one to the consumers and one to the source owner. 10. Log entry: draft it for the incident log, and write it after the yes. 11. Give the result as one web page. ## Output One HTML file, like [references/incident.html](references/incident.html): the title with the severity, the layers with the failed task, where, when and the status, the error message, then one section per step. Read it for the layout before you write the page. ## Reads - the run records, the check results, the quarantine, the delivered files, the pipeline code, the incident log ## Writes - the web page, and one entry in the incident log after the yes ## Never - Never fixes anything, never reruns a job, never sends an email.
© Data With Baraa · MIT · DataWithBaraa/claude-code-data-engineering-skills
Similar skills
All skills for DevOps and infrastructure →incident by other authors comes first.