NarekManukyan

Flutter Mobx Kit — Testing skill for Claude Code

Testing community

Agent toolkit for Flutter + MobX: AGENTS.md as the single instruction source, 15 on-demand playbooks, 17 ADRs, a plan-first delivery loop and a three-tier testing contract.

How to install Flutter Mobx Kit

This entry records only its repository, not the path inside it, so there is no exact command to give. Open NarekManukyan/flutter-mobx-kit and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Flutter Mobx Kit does

Agent toolkit for Flutter + MobX: AGENTS.md as the single instruction source, 15 on-demand playbooks, 17 ADRs, a plan-first delivery loop and a three-tier testing contract. Works with Claude Code, Codex, Cursor, Copilot and Gemini.

Alternatives in Testing

  • Deslop Reborn — One iteration of the IronClaw Reborn de-slop loop — take ONE Reborn crate, fan out parallel review sub-agents 12.6k ★
  • Deliver — Delivery phase - Review, validate, and test with multi-AI quality assurance 2.8k ★
  • GAAI Framework — 95+ Drop-in governance layer (.gaai/ folder) -- backlog authorization, cross-session memory, decision tracking 113 ★

README

flutter-mobx-kit

Agent toolkit for Flutter + MobX projects. One instruction source, playbooks that load only when they are needed, a delivery loop that reads the acceptance criteria before writing code, and a definition of done a machine can check.

Works with **Claude Code, Codex, Cursor, Copilot, Gemini, Zed, Amp** — anything that reads `AGENTS.md`.

curl -fsSL https://raw.githubusercontent.com/NarekManukyan/flutter-mobx-kit/main/install-remote.sh | bash

The problem it solves

Two things go wrong on a Flutter team using coding agents.

**Instructions get expensive and get ignored.** A `CLAUDE.md` grows to 400+ lines because every convention has to live somewhere. It is injected on every turn, so a session that renames a constant pays for the DTO authoring guide — and the rules that matter for the current task get no more attention than the ones that do not. Meanwhile the same rules are duplicated into `.cursor/rules`, and absent entirely for Codex and Copilot, so a developer who switches editors gets a different architecture.

**Agents build the wrong thing, then prove it works.** The expensive failures are not bad Dart — linters catch that. They are a screen that satisfies the ticket title but misses three acceptance criteria buried in the description, shipped with a green unit suite and no idea what happens when the server returns a 500.

This kit is the fix for both, extracted from a production boilerplate.

What you get

Three layers instead of one file

Layer Answers Where Loaded
AGENTS.md what the rules are repo root, ~200 lines always
Playbooks how to build a thing .claude/skills/{name}/SKILL.md on demand
ADRs why the rule exists docs/adr/ when you want to challenge a rule

`AGENTS.md` is the single source. `CLAUDE.md` (an `@AGENTS.md` import plus Claude-only wiring), `GEMINI.md`, `.github/copilot-instructions.md` and `.cursor/rules/000-agents.md