The Analyst Engineering Agent Kit -- A Self-Improving Agent, Skills, Hooks and MCP for Technical Analysts
You open a 60-page OpenAPI contract on a Monday. By Friday you are still hand writing user stories, copying field constraints into acceptance criteria, building a Postman collection by clicking, and pasting test results into Xray one row at a time. The work is mechanical, but it eats your week and it is where mistakes creep in.This kit removes the mechanical part and leaves you the judgment.It is not a prompt pack. It is an engineered system: a single self-improving agent, three working skills, hooks, an MCP configuration for Jira and Confluence, and adapters for both Claude Code and VS Code with GitHub Copilot. The intelligence lives in portable Python and Markdown, so every skill runs standalone from your terminal even if your agent setup is not wired up yet.What makes this different. It was built by a technical business analyst working on a payments API program at a bank, not by a content marketer. The examples are payments examples. The Xray updater exists because pushing test executions by hand is a real Friday afternoon problem. The "self-improving" claim is concrete and you can watch it work: the agent maintains a house-style memory file that the engines read on every run, so when you correct its output once, every future run follows your convention.What is insideThe mono-agent. One agent, sequential, never parallel. It reads your house style, applies it, and persists what it learns. AGENT.md defines the operating rules and the self-learning protocol in plain language you can edit.Skill 1: Contract to Artifacts. Point it at an OpenAPI or AsyncAPI contract plus an optional use-case file. It produces user stories, Gherkin acceptance criteria, functional rules, test scenarios, and a traceability matrix. Every rule traces to a field or response code in the contract.Skill 2: Postman and Bruno Builder. Generates daisy-chained, runnable collections from a contract. A create call captures an id into a variable and later calls reuse it. Ships both Postman JSON and git-native Bruno files so test changes are reviewable like code.Skill 3: Xray Test Execution Updater. Pushes results to an Xray Test Execution in Jira. Dry-run by default, so you review the exact payload before anything is written.Jira and Confluence via MCP. A ready MCP configuration so the agent can read tickets and pages as context and write specs back.Two runtime adapters. Claude Code with hooks and settings, and VS Code with GitHub Copilot. The skills are identical across both. Only the orchestration glue differs.The setup and usage guide. A 13-page dark-themed PDF that walks you from install to a full worked example end to end.Who this is for Technical business analysts who write requirements and tests from contracts BAs and QA analysts on API, banking, or payments programs Analysts who already live in Claude Code or VS Code and want real leverage Anyone tired of prompt packs who wants a system that actually runs Who this is NOT for People looking for a one-click no-code tool. This is for analysts comfortable running a Python command. People who want generated requirements they never review. The agent marks assumptions for a reason. FormatDigital download. PDF guide (13 pages) plus a complete repo: agent, three skills with their own SKILL.md files, MCP config, two adapters, and a sample contract and use case to run immediately. AI access path is your GitHub Copilot license. No separate API tokens required.By Ahmed Squalli Houssaini. Part of the Analyst Engineering ecosystem at analystengineering.com.
Get it → ahmedsqualli.gumroad.com