sentinel-guard

# Sentinel Guard Lightweight file access monitor that protects sensitive files from AI coding agents. Watches for `.env`, API keys, SSH keys, crypto wallets, and other secrets being accessed by tools like OpenClaw, Cursor, or Copilot. Logs everything, alerts in real-time, and can block access via agent hooks. ## Install ```bash cargo install --path . ``` Or download the binary from [Releases](https://github.com/Supawitk/sentinel-guard/releases). ## Commands ### Setup ```bash sentinel init # Generate config file sentinel check # Verify installation ``` ### Scanning ```bash sentinel scan . # Find sensitive files sentinel scan . --deep # + check file contents for secrets sentinel skill-scan . # Scan AI agent skills for malicious patterns sentinel skill-scan --file X.md # Scan a specific skill file ``` ### Monitoring ```bash sentinel watch . # Real-time file access monitoring sentinel hook # Run as AI agent PreToolUse hook (stdin/stdout JSON) ``` ### Integrity ```bash sentinel baseline . # Hash sensitive files sentinel verify . # Check for changes since baseline ``` ### Reporting ```bash sentinel log # View activity log sentinel log --sensitive # Sensitive events only sentinel stats # Event statistics sentinel dashboard # Interactive TUI dashboard sentinel report . -o report.csv # Export to CSV sentinel report . -t log -o x.csv # Export activity log ``` ### Maintenance ```bash sentinel cleanup --days 7 # Remove old log entries ``` ## Config Generate with `sentinel init`, customize `sentinel.toml`: ```toml [watch] paths = ["."] recursive = true [protect] sensitive_patterns = ["**/.env", "**/*.pem", "**/id_rsa", "**/wallet.dat"] [alert] mode = "all" # "log", "terminal", "all" action = "warn" # "warn" or "block" desktop_notifications = false [log] db_path = "~/.sentinel-guard/activity.db" retention_days = 30 # Optional webhooks [[webhooks.endpoints]] name = "slack" url = "https://hooks.slack.com/services/..." format = "slack" # "slack", "discord", "generic" enabled = false ``` ## What It Detects **Files:** `.env`, SSH keys, certificates, AWS/GCloud configs, crypto wallets, OpenClaw secrets, package manager configs **Secrets in code:** API keys (AWS, OpenAI, Anthropic, GitHub, GitLab, Slack, Stripe), private keys, database URLs, crypto wallet keys, seed phrases **Malicious AI skills:** Data exfiltration commands, RCE patterns, credential harvesting, persistence mechanisms, browser data access ## Architecture ``` src/ ├── core/ Config, database, shared types ├── detect/ Secret scanner, skill scanner, integrity checker ├── monitor/ File watcher, rules engine, agent hooks ├── output/ Alerts, dashboard, notifications, webhooks, reports └── main.rs CLI commands ``` ## License MIT