fujahgabriel

PIIGuardAI — AI skill for Claude Code

AI community

A macOS menu bar app that blocks messages containing PII before they reach Claude, ChatGPT, Gemini, and other LLM providers -- from both browsers and command-line tools (IDE agents, curl, scripts).

How to install PIIGuardAI

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

What PIIGuardAI does

A macOS menu bar app that blocks messages containing PII before they reach Claude, ChatGPT, Gemini, and other LLM providers -- from both browsers and command-line tools (IDE agents, curl, scripts).

Alternatives in AI

Skill Seekers Roadmap — Transform Skill Seekers into the easiest way to create Claude AI skills from any knowledge source - documentat 11.1k ★
Manifest — Real-time cost observability for OpenClaw agents — track tokens, costs, messages, and model usage 4.1k ★
Browserwing — BrowserWing turns your browser actions into MCP commands Or Claude Skill, allowing AI agents to control browse 1.2k ★

README

PIIGuard AI

A macOS menu bar app that blocks messages containing PII before they reach Claude, ChatGPT, Gemini, and other LLM providers -- from both browsers and command-line tools (IDE agents, `curl`, scripts).

How it works

Local root CA. On first "Turn on protection", the app generates a self-signed root certificate authority (P-256, via Apple's swift-certificates) and asks you (one admin-password prompt) to trust it in the System keychain.
Scoped interception.
- Browsers: a PAC (Proxy Auto-Config) file routes only the known LLM provider domains (claude.ai, api.anthropic.com, chatgpt.com, api.openai.com, generativelanguage.googleapis.com, etc. -- see Providers tab) through a local proxy on 127.0.0.1. Everything else on your Mac goes DIRECT and is never touched.
- CLI tools: browsers honor the system PAC automatically, but curl, Node, Python, and most CLI HTTP clients don't. The app auto-writes a sourceable env file (HTTP_PROXY/HTTPS_PROXY + NODE_EXTRA_CA_CERTS/ SSL_CERT_FILE/CURL_CA_BUNDLE/REQUESTS_CA_BUNDLE pointing at a PEM copy of the CA) and installs a one-line source into your shell rc files, plus sets the same vars via launchctl for GUI apps like VS Code. See the "CLI protection" section in Settings > General. Open a new terminal window after toggling -- existing shells keep whatever env they already exported.
Local MITM proxy. For each intercepted connection, the app terminates TLS using a certificate it mints on the fly for that hostname (signed by the root CA from step 1), reads the decrypted HTTP request, and:
- runs it through a local, on-device PII scanner (email, phone, SSN, credit card w/ Luhn check, AWS keys, generic API keys/secrets, IPv4, .env-style secrets/files) plus any custom keyword/regex rules you add
- if PII is found, blocks it (returns an HTTP 403 to the app/browser