If you discover a security vulnerability in IntelShed, please report it responsibly:

1. Do not open a public GitHub issue.
2. Use GitHub Security Advisories (preferred) or email the maintainer.
3. Include: description, reproduction steps, affected versions.
4. Response time: within 48 hours.
5. Coordinated disclosure after fix is released.

• IntelShed backend (FastAPI, 281+ routes)
• IntelShed frontend (React, Cesium)
• Docker stack (Caddy, PostgreSQL, Redis, Tor)
• Mesh protocol (node ingest, delta sync, Ed25519)
• MCP server (77+ tools)

• Third-party dependencies (report upstream)
• Self-hosted instances with modified security configuration
• OSINT data sources (report to the source provider)

IntelShed implements defense-in-depth:

• 4-layer anti-hallucination — prompt guard, RAG integrity, session guard, output guard
• RBAC + JWT + WebAuthn passkeys — role-based access control with hardware-backed authentication
• CSP single source of truth — csp_policy.py feeds header, meta tag, and Caddyfile
• Rate limiting — Redis sliding window with in-memory deque fallback
• HMAC + Ed25519 node authentication — mesh nodes sign all telemetry and briefing payloads
• Zero-Trust Ingest PEP — hash-chained evidence capture with W3C PROV DAG
• Secret management — scoped API keys, .env isolation, no hardcoded secrets
• Pre-commit secret guard — .husky hook scans for accidental secret commits

• Set WORLDBASE_API_KEY in backend/.env (random 32+ chars)
• Set WORLDBASE_NODE_INGEST_TOKEN for mesh authentication
• Review WORLDBASE_CORS_ORIGINS — do not use * in production
• Enable WORLDBASE_REQUIRE_NODE_TOKEN=1 (default)
• Review Caddyfile @blocked directive for LAN subnet restrictions
• Do not expose backend port 8002 directly — use Caddy proxy
• Review WORLDBASE_MCP_WRITE and WORLDBASE_AGENT_BUS for LAN-exposed instances