Keep your engineering skills relevant as AI changes the industry around you.

Production Investigation Kit

Evidence-first production investigations with Claude Code, on top of scoped, least-privilege Azure access. The security design is part of the resource: an investigator-only identity, restricted SQL views, auditing and separation of duties, with agent-side guardrails layered on top.

• Investigator sub-agent plus skills for the investigation procedure, SQL queries and KQL / Log Analytics queries
• Azure identity, group and role setup, investigation views and read-only database access
• Auditing, and a separation between investigator access and audit-log access
• A guard hook that blocks write operations, Azure changes and account switching
• PowerShell and Bash setup scripts, prerequisite checks, guard tests and a verification checklist
• Worked examples and an evidence-linked report structure

Production_Investigation_Kit/
├── .claude/
│   ├── agents/investigator.md
│   ├── hooks/investigation_guard.py
│   └── skills/  investigate · query-sql · query-kql
├── azure/       01-create-groups … 07-verify-access
├── GUIDES/      01_WHAT_YOU_WILL_BUILD … 06_CHECKLIST
├── tools/       check_prereqs · scan_for_identifiers · test_guard
└── Production_Investigation_Kit_Guide.pdf

Assumes Claude Code, Azure App Service, Azure SQL and Application Insights / Log Analytics. It documents an opinionated, working setup: adapt it, test it in a sandbox first, and never run it directly against production.

“I especially like that it goes beyond the concepts and covers the actual implementation, from setting up least-privilege access and auditing to integrating Claude Code and adapting the workflow to a real system. This kind of practical, security-conscious approach to AI-assisted engineering is exactly what I'd like to learn more about.”

Amirhosein, .NET Developer

Technology Health & Unit Economics Review Toolkit

A repeatable periodic review that looks at system health and economic sustainability together: cost to run, unit economics, scalability, delivery performance, reliability, security posture and technical risk, with a trend against the previous review.

• A blank workbook and a privacy-screened Atherio example workbook (June to August 2026)
• A standard-library Python generator that turns a workbook into an HTML report
• Optional comparison against the previous period by matched metric keys
• A metric schema and a guide to reading metrics responsibly

Technology_Health_Unit_Economics_Review_Toolkit/
├── example-atherio/  Atherio_Public_Example.xlsx · _Report.html
├── generator/        generate_report.py · metric_schema.json
├── guide/            Technology_Health_Review_Guide.pdf
└── templates/        Technology_Health_Blank.xlsx · Blank_Report_Preview.html

Not connected to billing or telemetry providers: the accuracy of a review depends on the evidence you enter.