Agentic Defaults — Testing skill for Claude Code

agentic-defaults

Sensible defaults for agentic coding: test-first, reasonable coverage, and context that stays current.

Generic instructions, skills and agents for AI coding assistants, built around practical delivery and modest model budgets. Everything in [`kit/`](kit/) works in both **Claude Code** and **GitHub Copilot** (VS Code). Other tools that read `AGENTS.md`, such as Cursor, get the working agreement; the skills and agents have only been tested in Claude Code and Copilot. Install it in a project with [agd](https://github.com/Zejnilovic/agd), or copy what you need.

**Why it works this way:** [Working with coding agents](https://zejnilovic.eu/guides/working-with-coding-agents) explains the method behind the kit ([same text in this repo](docs/methodology.md)). It covers the principles, rules of thumb, anti-patterns and nine hands-on exercises, and is written to stay true as tools and models change.

Quick start

1. **Install** [agd](https://zejnilovic.github.io/agd), with Homebrew:

   brew install zejnilovic/tap/agd

   Or without Homebrew:

   curl -fsSL https://raw.githubusercontent.com/Zejnilovic/agd/main/install.sh | bash

2. **Add the kit** in your project's root, then commit the new files together with `.agd.lock`:

   agd init

   That installs every [pack](#packs). To install only some, name them, for example `agd init --with flask,react`; the packs they require come along. Add `copilot-hook` to the list if you use Copilot. Without agd, copy the contents of `kit/` into the project's root.

3. **Scan** the project: ask your assistant to "use the analyst to scan this project". It writes the root `CONTEXT.md`, proposes a [project profile](#project-profile), fills in the [standard commands](#standard-commands) in `AGENTS.md`, and proposes any that are missing. Review and confirm them.

4. **Work.** For small changes, just chat; the working agreement applies. For bigger ones, give it the ticket n