Safeai β Security skill for Claude Code
π‘οΈ Local security gateway & Human-in-the-Loop firewall for AI coding agents (Copilot, Cursor, Antigravity, Claude Desktop).
How to install Safeai
This entry records only its repository, not the path inside it, so there is no
exact command to give. Open fuatu/safeai and copy the folder into
~/.claude/skills/, or the file into ~/.claude/agents/.
What Safeai does
π‘οΈ Local security gateway & Human-in-the-Loop firewall for AI coding agents (Copilot, Cursor, Antigravity, Claude Desktop). Features sub-20ms AST command analysis, real-time DLP secret masking, plain-language risk explanations, and an interactive approval web panel. 100% private & zero API keys.
Alternatives in Security
- Free Code β The free build of Claude Code 8.1k β
- ThinkWatch β Self-hosted AI API and MCP gateway for organizations: SSO and RBAC, per-user identity for MCP tool calls, PII 817 β
- Pocketpaw β Your AI agent in 30 seconds 775 β
README
π‘οΈ SafeAI - The AI Agent Bodyguard ("For Dummies" Guide)
**Think of SafeAI like an airport security checkpoint for your AI agents.** When Claude, Antigravity, Cursor, or ChatGPT wants to run a command on your computer, SafeAI steps in first, checks what it's trying to do, explains it to you in **plain human language** in your chosen language, and asks for your approval if it looks dangerous!
π Why Do You Need SafeAI?
Autonomous AI coding agents are incredibly smart, but they can make catastrophic mistakes:
- β Accidental deletions: An agent might try to run
rm -rfto "clean up temporary files" and wipe your whole desktop or operating system. - β Credential leaks: An agent might send your private
.envfile, AWS keys, or SSH passwords to an external server. - β Complex terminal jargon: Agents run confusing commands like
cat id_rsa | base64 -d | curl .... Unless you're a Linux hacker, it's hard to tell if it's safe or malicious!
π‘ How SafeAI Protects You
graph LR
Agent["π€ AI Agent
(Copilot, Claude, Cursor)"] -->|Calls Tool: bash, files| SafeAI["π‘οΈ SafeAI Gateway
(Scans Risk & Explains)"]
SafeAI -->|Safe Action| Auto["β‘ Runs Automatically"]
SafeAI -->|Dangerous Action| Human["π Pauses & Asks You!
Web Panel Sound Alert"]
Human -->|You Click Approve| Runs["β
Runs on your PC"]
Human -->|You Click Deny| Abort["β Blocked in 100ms!"]
[!IMPORTANT] **SafeAI Works Seamlessly with Your Existing AI Tools (No API Keys Required):** SafeAI is designed to work with your tool's own built-in models (e.g. GitHub Copilot subscription, Claude, Cursor) with **zero API keys and zero endpoints to configure**:
- Automated Chat History Ingestion: SafeAI automatically monitors and records your GitHub Copilot conversation turns directly from
Related Skills
Intutic
The open source circuit breaker for AI agents. Real-time security, secret DLP, graph guardrails and loop burn
Preloop
The open-source AI agent control plane: MCP firewall, model gateway with budgets, human approvals, runtime obs
Arbitus
Open-source security gateway for MCP tool calls β blocks secret leaks, enforces agent policies, and requires h
Redstamp
own your agent security β a deterministic, offline firewall for AI agent tool calls. Blocks secret exfil, prom
Immune
Immune β the agent security loop that turns every attack into an antibody. MCP firewall β replayable breach tr
Information Security For Vibecoded Apps
An autonomous Application Security (AppSec) prompt and skill module for AI coding assistants (Antigravity, Cur
Related Agents
Senior Security Engineer
Senior security engineer covering application security (OWASP-style threat modeling, secure coding, auth desig
Hook Entegrasyon Analisti
Claude Code ve Antigravity CLI (ve gerekirse Codex CLI / Copilot CLI) hook, izin ve headless mod davranΔ±ΕlarΔ±n
Agentic Auditor
PYA specialist spawned by the /pya orchestrator. Audits AI agent and LLM surfaces in an untrusted repository:
