firaskudsy

Se Forge — Testing skill for Claude Code

Testing community

A team of AI agents in Claude Code that builds client apps from requirements to merged, tested code: discovery, stories, architecture and API contract, then plan, build, review, QA and merge per ticke.

How to install Se Forge

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

What Se Forge does

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 ★
Test Scenarios — Create comprehensive test scenarios from user stories 7.8k ★
Cc Sdd — Spec-driven development (SDD) for your team's workflow 3.1k ★

README

se-forge

**A team of AI agents in Claude Code that builds client apps from requirements to merged, tested code, under safety rules it cannot talk its way around.** You give it the client's requirements and screens; it asks you what they leave open, writes the stories, the architecture, the screens and the API contract, cuts the work into tickets, and then plans, builds, reviews, tests and merges them one by one, while you make the decisions. It can also work in an existing system of several repositories.

A team, not one chatbot. 22 agents with one job each: a planner, analysts, a designer and an architect; database, backend, frontend, mobile and devops engineers; code, security, contract, database, infrastructure, UI and performance reviewers; QA; a project manager; and a lead that runs them.
Rules that hold. A safety layer of Claude Code hooks checks every command and file write before it runs: no pushes to protected branches, no secrets read or committed, no "done" without green tests, no merge without passing reviews, and the risky steps wait for a person.
Production quality by default. Contract-first APIs (OpenAPI 3.1), tests on every side and end-to-end flows, a security review against OWASP ASVS, decision records, containers and CI from day one.
Nothing in the client's repository. The kit lives beside the code, hidden from git; plans, reviews and settings go to a private team repository.
Decisions in a few keystrokes. Whatever needs you comes as choices: pick one (A, B, C), tick any ([ ]), yes or no, or your own words when nothing fits. You reply 1A 2B, not an essay.
You see where it goes. Every step is recorded: a local dashboard shows time, waits for you, tokens and cost per ticket and per agent, the architecture, the screens, and a forecast of the finish.

Version 0.2.41. Built for Claude Code on Claude subscriptions, on macOS, with GitLab or GitHub. New here? Read [GUIDE.md](GUIDE.md): step by step,