JustinLaSalle

Agentic Codeforge — Security skill for Claude Code

Security community

Enterprise code generation and review agent — writes new software from a plain-language spec (with live sandboxed preview for web widgets), or reviews existing code for real bugs, security vulnerabili.

How to install Agentic Codeforge

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

What Agentic Codeforge does

Alternatives in Security

Security Best Practices — Review code for language-specific security vulnerabilities 14.6k ★
Roast Repo — Deliver a brutally honest, technically sharp roast of the current repository 186 ★
Omarchy Plugin Security Skill — Agent skill and field guide: the security pitfalls that block Omarchy plugins on the marketplace, from 5,000+ 70 ★

README

CodeForge

An enterprise code generation and review agent — writes new custom software from a plain-language spec, or reviews existing code for real bugs, security vulnerabilities, and performance issues, then produces a corrected version. Built with the [Claude API](https://docs.claude.com).

Live demo

**[Try it here](https://claude.ai/artifact/PoSnXUMsyyphygEc5z1ffV)** — note: trying the interactive part requires a free Claude account to sign in with (a platform requirement for any page that calls Claude live, not specific to this project).

The problem

Two of the most common software requests in an enterprise setting are "build us something custom for this workflow" and "this existing code has problems, can someone look at it." Both normally require a developer's time just to produce a first pass. This project automates a genuinely useful first pass at both — not to replace engineering judgment, but to get a working starting point or a real, specific list of problems much faster than starting from a blank page or a cold read-through.

What it does

**Build New** — describe a requirement in plain language and choose an output type:

Interactive web widget — generates a complete, self-contained HTML/CSS/JS widget and renders it live, immediately, in a sandboxed preview right in the page. It's genuinely functional, not a mockup — an invoice calculator example actually adds/removes line items and computes totals in real time.
Code snippet — for any other language (Python, JavaScript, Java, C#, Go, SQL, TypeScript), writes idiomatic, commented code to the spec.

**Review & Improve** — paste existing code (optionally with context about where it runs), and the agent:

Finds real, specific issues — bugs, security vulnerabilities (e.g. injection risks), performance problems, style violations — each tagged with severity and category, not generic nitpicks. It's explicitly instructed to report no issues if the code is genuinely solid, rather tha