Post Mortem

Post-Mortem Skill

Structured failure analysis workflow for diagnosing and learning from workflow failures, build breaks, and operational incidents.

Overview

This skill provides a systematic approach to failure analysis when something goes wrong in the AWK workflow pipeline. It walks through evidence collection, root cause identification, impact assessment, and actionable remediation -- producing a structured report that can be referenced for future prevention.

When to Use

Use this skill when:

• A Worker dispatch fails or times out
• A PR is rejected due to unexpected errors
• CI/CD pipeline breaks
• awkit kickoff encounters unrecoverable errors
• Any workflow step produces unexpected results
• User invokes /post-mortem

Workflow

Phase 1: Analyze Failure

Read phases/analyze-failure.md

Walk through the structured failure analysis:

1. Collect failure context (trace files, logs, error messages)
2. Identify root cause (categorize: config, code, environment, external)
3. Determine blast radius (what was affected)
4. Generate remediation steps
5. Create prevention measures
6. Output structured report

Critical Rules

1. DO NOT modify any files during analysis -- this is a read-only diagnostic skill
2. Collect evidence first -- never jump to conclusions without supporting data
3. Categorize accurately -- misclassification leads to wrong remediation
4. Be specific in remediation -- vague advice is not actionable
5. Always produce the structured report -- even if root cause is uncertain

Integration with AWK Workflow

This skill is used after a failure has occurred. It does not interfere with active workflows.

• Safe to run while awkit kickoff is active (read-only)
• Results can inform new issues created via /create-issues
• Prevention measures may lead to config or rule updates

Self-Check

On each phase entry, output:

[POST-MORTEM] <timestamp> | <phase> | loaded: <filename>

Quick Reference