ROCCO BUTTIGLIERI · INDEPENDENT SOFTWARE BUILDER

I build practical software where AI, security, and automation meet.

Star Davos Software turns messy technical problems into focused tools: privacy-first analyzers, ChatGPT integrations, developer utilities, secure prototypes, and automation experiments.

Explore the work View GitHub

SELECTED WORK / 01

Software built around real constraints.

No fake dashboards. No invented metrics. The strongest projects here are designed around actual security, privacy, reliability, and usability problems.

FLAGSHIP · LOCAL-FIRST SECURITY TOOLING

Code Safety Analyzer

A local-first analyzer for scripts, projects, and files before they are run or shared with an AI service. It treats submissions as inert data, separates code risk from AI-upload risk, and produces evidence-linked explanations instead of pretending static analysis is certainty.

PowerShell, Python, Bash, and Windows Batch/CMD analysis
Privacy scanning and redaction workflows
Bounded cross-file reasoning
Evidence-based risk language and explicit uncertainty

Project repository →

ANALYSIS / LOCAL

CODE RISKMODERATE

AI UPLOAD RISKHIGH

clients.csv→Get-Content→Invoke-RestMethod

CMD-004FLOW-002CORR-001

Review what leaves the machine before execution or upload.

deploy.sh1,500 lines · copy verified

#!/usr/bin/env bash
set -euo pipefail

# Complete source stays intact.
# Copy stays locked until integrity matches.

validate_payload() {
  local expected="$1"
  local actual="$2"
  [[ "$expected" == "$actual" ]]
}

# Source is inert text.
...

INTEGRITY MATCH

CHATGPT / MCP · DEVELOPER UX

Compact Code Viewer

A ChatGPT MCP App for one specific annoyance: large generated code blocks swallowing the conversation. It keeps complete source in a bounded viewer and refuses to unlock copy when final integrity metadata does not match.

Rejects oversized input instead of silently truncating
Never executes submitted source
Docker hardening and adversarial test coverage
Cloudflare-hosted public release architecture

View public project →

EXPERIMENT · AUTOMATION + BUSINESS SYSTEMS

Project Zero

An experiment in building automated revenue-generating assets from a tightly constrained starting point. The interesting part is the operating discipline: preserve capital, record decisions, keep human approval on consequential actions, and preserve failed experiments instead of rewriting history.

$50 initial capital0 hidden funding100% decisions recorded

View project foundation →

CAPABILITIES / 02

Useful work I can do for a business.

Solve a narrow expensive problem, keep the architecture understandable, and make the result maintainable after the demo.

AI integrations

ChatGPT apps, MCP tools, controlled AI-assisted workflows, and interfaces that make model output easier to use safely.

Internal tools

Purpose-built software for repetitive support, operations, security, or data-handling tasks that do not justify an oversized platform.

Automation

Workflow design that reduces manual steps while keeping approvals, logging, rollback, and failure modes visible.

Security-minded prototypes

Early products with sensible trust boundaries, least privilege, secret handling, input controls, and testable release gates from the beginning.

HOW I BUILD / 03

Fast is useful. Fast without evidence is just guessing.

Prefer the smallest architecture that safely solves the current problem.

Treat code that looks correct differently from behavior that has been tested.

Make security boundaries, privacy assumptions, and failure modes explicit.

Build for actual human use, then change what feels wrong.

STAR DAVOS SOFTWARE · NEW JERSEY / REMOTE

Have a weird technical problem that should probably be software?

That is the kind of conversation I want.

Start with GitHub Back to launch