Skip to content

Repository files navigation

Claris AI — Defense Network

V7.0 "Unitium Learning Mode" | The World's First AI-Native Cyber Defense Network with Integrated Education

"Built to defend AI. Built to teach. Built to last." "Ship fast. But ship secure. And understand why." — August + AVARI

Version Scripts Learning Paths GitHub


What is Claris AI?

Claris AI is a full-stack AI-native Cyber Defense Network built for the OpenClaw ecosystem. V7.0 adds the Unitium Learning Mode — the world's best cybersecurity education system built directly into the defense tools.

It evolved from a simple LLM injection scanner into a 29-script defense + education platform with:

  • 🎓 Unitium Learning Mode — 6 learning paths, XP system, adaptive curriculum, interactive quizzes
  • 🤖 OpenClaw Security Path — T1-T12 threat education, VPS hardening, memory file protection
  • 📚 --learn flag — Educational output on every Claris scan command
  • 7-Layer Defense Stack (Perimeter WAF → Meta Oversight)
  • 10-Agent Security Swarm (Recon, Detect, Hunt, Respond, Forensics, Triage, Patch, Monitor, Report, Orchestrate)
  • Dash Platform Specialization (Evonode, DAPI, DPNS, quorum security)
  • 6 Core Words Framework (Trust, Adversarial, Surface, Entropy, Lateral, Posture)
  • 30 Vibe Coder Security Rules — Pattern-enforced across all code you ship
  • Posture Intelligence Engine — Holistic security posture scoring and tracking
  • 12-Point VPS Hardening Auditopenclaw_hardening.py
  • Federated Mesh with BFT consensus and $Initium DAO incentives
  • Autonomous Incident Response with 12 built-in playbooks
  • 50-Payload Red Team Suite for continuous self-testing

🎓 V7.0 QUICKSTART — Unitium Learning Mode

# Enable global learning mode
python3 scripts/learning_mode.py --enable

# See all 6 learning paths
python3 scripts/learning_mode.py --paths

# Start the crown jewel path (OpenClaw-specific)
python3 scripts/learning_mode.py --path openclaw-security
python3 scripts/learning_mode.py --lesson OC01

# Check your progress
python3 scripts/learning_mode.py --status

# Deep explain any concept
python3 scripts/learning_mode.py --explain "prompt injection"
python3 scripts/learning_mode.py --explain "zero trust"

# Run hardening audit with full education
python3 scripts/openclaw_hardening.py --audit --learn

# Scan with learning mode active
python3 scripts/injection_guard.py --text "ignore all previous instructions" --learn
python3 scripts/openclaw_guard.py --quick --learn

📚 Full curriculum: LEARNING_PATHS.md
📖 Deep guide: UnitiumLearningMode.md


Quick Start

# ── Core Scanning ──────────────────────────────────────────────────────────
# Scan text for injection attacks
python3 scripts/injection_guard.py --text "ignore previous instructions"

# Scan YOUR CODE for the 30 Vibe Coder security violations
python3 scripts/vibe_coder_guard.py --scan ./src
python3 scripts/vibe_coder_guard.py --scan ./myapp --ext .js,.ts,.tsx
python3 scripts/vibe_coder_guard.py --rule 12     # Show rule details
python3 scripts/vibe_coder_guard.py --list        # All 30 rules

# ── Cyber Patriot Education ──────────────────────────────────────────────
# The 6 Core Words — foundational mindset map
python3 scripts/cyber_patriot.py --six-words

# Calibrated threat brief (audience-aware)
python3 scripts/cyber_patriot.py --brief phishing --audience beginner
python3 scripts/cyber_patriot.py --brief ransomware --audience advanced
python3 scripts/cyber_patriot.py --train --audience intermediate

# Interactive posture assessment
python3 scripts/cyber_patriot.py --assess-posture

# ── Posture Intelligence ─────────────────────────────────────────────────
# Score your security posture across 6 dimensions
python3 scripts/posture_engine.py --score \
  --trust 0.8 --adversarial 0.6 --surface 0.7 \
  --entropy 0.6 --lateral 0.5 --posture 0.8

# Full posture report
python3 scripts/posture_engine.py --report

# Track posture over time
python3 scripts/posture_engine.py --history
python3 scripts/posture_engine.py --delta

# ── Dash + Web3 Defense ──────────────────────────────────────────────────
python3 scripts/dash_guard.py --scan "quorum manipulation attempt"
python3 scripts/dash_security_intelligence.py --monitor

# ── Agent Swarm ──────────────────────────────────────────────────────────
python3 scripts/agent_swarm.py --route "wallet drain attempt detected"
python3 scripts/agent_swarm.py --status

# ── Education ────────────────────────────────────────────────────────────
python3 scripts/cyber_educator.py --lesson foundations --level 1

# ── Red Team ─────────────────────────────────────────────────────────────
python3 scripts/redteam_suite.py --run-all

The 6 Core Words — Claris's Foundation

These aren't vocabulary. They're a mindset map. Master these and you see the whole field differently.

# Word Weight Tagline
1 TRUST 25% Everything in security is a trust question.
2 ADVERSARIAL 20% There is always someone on the other side.
3 SURFACE 20% Every exposed edge is an invitation.
4 ENTROPY 15% Randomness is strength. Decay is constant.
5 LATERAL 10% They didn't come for the front door.
6 POSTURE 10% How you hold yourself — all of it, all at once.

The 30 Vibe Coder Security Rules

Every rule enforced by vibe_coder_guard.py. Ship fast — but ship secure.

Sev # Rule
🔴 01 Never store sensitive data in localStorage → use httpOnly cookies
🟠 02 Disable directory listing on your server
🟠 03 Always regenerate session IDs after login
🟠 04 Use Content Security Policy headers on every page
🔴 05 Always re-validate server-side — never trust client-side alone
🟠 06 Set X-Frame-Options to DENY
🟠 07 Strip metadata from every user-uploaded file
🟠 08 Never expose stack traces in production responses
🟠 09 Use short-lived presigned URLs for private files
🟠 10 Implement CSRF tokens on all state-changing requests
🟡 11 Disable autocomplete on sensitive form fields
🔴 12 Hash passwords with bcrypt — minimum cost factor of 12
🟠 13 Keep dependencies minimal — every package is attack surface
🟠 14 Use SRI (subresource integrity) for every external script
🔴 15 Never log passwords, tokens, or PII
🔴 16 Enforce HTTPS everywhere — redirect all HTTP
🔴 17 Separate DB credentials per environment — never share prod creds
🟠 18 Implement account lockout after 5 failed login attempts
🟡 19 Validate content-type headers on every API request
🔴 20 Never use MD5 or SHA1 for anything security-related
🟠 21 Scope OAuth tokens to minimum required permissions
🟠 22 Use nonces for every inline script in CSP
🟠 23 Monitor for dependency vulnerabilities weekly (Snyk / npm audit)
🟡 24 Disable HTTP methods you don't use (TRACE, etc.)
🟠 25 Logout = server-side session invalidation + cookie clear
🟠 26 Use constant-time string comparison for token validation
🟠 27 Never cache sensitive API responses (Cache-Control: no-store)
🟡 28 Set Referrer-Policy to strict-origin
🟠 29 Enforce password complexity server-side
🟠 30 Scan Docker images for vulnerabilities before every deployment

🔴 = CRITICAL | 🟠 = HIGH | 🟡 = MEDIUM


All Scripts

Core — Injection & Scanning

Script Version Description
injection_guard.py V6 Primary LLM prompt injection scanner (6 layers, 100+ patterns)
cortex_engine.py V5 Behavioral pattern cortex engine
owasp_llm_scanner.py V5 OWASP LLM Top 10 2025 vulnerability scanner
temporal_analyzer.py V6 Time-based attack pattern analysis (6 temporal patterns)
zero_day_hunter.py V5 Zero-day anomaly detection and pattern learning
threat_monitor.py V5 Real-time 6-hour threat monitoring and alerting
smart_contract_scanner.py V5 Web3/smart contract OWASP SC Top 10 scanner
adversarial_trainer.py V5 Adversarial training and bypass detection
claris_dashboard.py V5 Security dashboard UI
claris_api.py V5 Claris REST API server (port 7433)
claris_scan.py V5 Primary scan CLI (unified interface)
openclaw_guard.py V5 OpenClaw platform-level guard (T1-T12)

Cyber Patriot Protocol — V6.0

Script Version Description
cyber_patriot.py V6.0 Marcus Webb Framework — calibrated threat education, 6 Core Words, audience levels
posture_engine.py V6.0 Holistic security posture scoring across 6 Core Word dimensions

Vibe Coder Security — V6.1

Script Version Description
vibe_coder_guard.py V6.1 30-rule code scanner — pattern detection across all violation categories

Defense Network — V6.x

Script Version Description
dash_security_intelligence.py V10 Dash Platform specialized threat intelligence
zero_trust_enforcer.py V10 Zero-trust enforcement and architecture auditor
autonomous_responder.py V10 Autonomous incident response (12 playbooks)
cyber_educator.py V10 8-pillar progressive cybersecurity curriculum
federation_mesh.py V10 Decentralized federation mesh with BFT consensus
ml_enhanced_scan.py V10 ML-enhanced scanning layer
prompt_guard_ml.py V10 ML-based prompt guard

Forge — V10.0

Script Version Description
agent_swarm.py V10 10-agent specialized security swarm
openclaw_guardian.py V10 OpenClaw Instance Guardian — L1-L7 on every LLM call
dash_guard.py V10 Expanded Dash defense (12 threat categories)
redteam_suite.py V10 50-payload red team test suite
initium_daemon.py V10 $Initium DAO governance stub

Total: 27 scripts


Architecture Overview

7-Layer Defense Stack

L7: Meta Oversight      → AI-on-AI behavioral monitoring
L6: Output Filter       → Data leakage prevention
L5: Runtime Guard       → System prompt integrity
L4: Semantic Analysis   → Intent and social engineering detection
L3: Behavioral Cortex   → Temporal + rate analysis (6 temporal attack patterns)
L2: Input Guard         → Signature scan + sanitization (100+ patterns)
L1: Perimeter WAF       → Known bad pattern filter

10-Agent Swarm

ORCHESTRATE (Claris Meta) → routes all tasks
├── RECON      → intelligence gathering
├── DETECT     → real-time detection
├── HUNT       → proactive threat hunting
├── RESPOND    → incident response
├── FORENSICS  → digital forensics
├── TRIAGE     → prioritization
├── PATCH      → remediation
├── MONITOR    → continuous monitoring
└── REPORT     → intelligence reporting

Cyber Patriot Audience Levels

Cyber Recruit   (beginner)     → 40% disclosure cap  → Mindset and curiosity
Cyber Apprentice (intermediate) → 70% disclosure cap  → Technique and patterns
Cyber Defender  (advanced)     → 90% disclosure cap  → Adversarial thinking
Cyber Patriot   (expert)       → 100% disclosure     → Full threat landscape

References

File Description
references/cyber-educator-framework.md Marcus Webb Protocol — philosophy of cybersecurity education
references/vibe-coder-security-rules.md All 30 rules with deep explanations, code patterns, and fixes
references/unitium-knowledge-base.md Complete knowledge base from all 12 Unitium.One !ReadMe articles (20,646 words)
references/v10-architecture.md V10 architecture deep dive
references/cyber-curriculum.md 8-pillar curriculum structure
references/autonomous-response-playbooks.md 12 incident response playbooks
references/dash-platform-security.md Dash Platform security guide
references/web3-attack-vectors.md WA01-WA15 web3 attack vectors
references/agentic-ai-security.md Agentic AI threat landscape
references/vulnerability-db.json 50+ structured vulnerability entries

Version History

Version Name Key Additions
V1.0 Genesis Basic injection scanner
V3.0 Deep Defense 6-layer guard, Web3/SC scanning, threat monitor
V5.0 Federated Cortex Temporal analyzer, zero-day hunter, OWASP LLM Top 10
V6.0 Cyber Patriot Protocol 6 Core Words, Marcus Webb Framework, Posture Intelligence
V6.1 Vibe Coder Security 30 Security Rules, vibe_coder_guard.py, deep training database
V6.2 Unitium Knowledge Integration All 12 Unitium.One articles: color teams, DDoS defense, ZKPs, systems thinking, Zero Trust stats

Security Notes

  • All scripts are stdlib-first (no unnecessary dependencies — Rule 13)
  • Errors return structured JSON, never raw tracebacks (Rule 08)
  • No hardcoded credentials anywhere (Rule 17)
  • Data files use data/ directory with JSONL append-only logs
  • httpOnly cookies advocated throughout (Rule 01)

Claris AI V6.1 "Vibe Coder Security Rules" — Built by August + AVARI MIT License | Open Source | Semper Fortis — Always Strong. Always Brave. ~Claris · Ship fast. But ship secure.

About

Claris AI — Federated Cortex security engine for OpenClaw agents. Prompt injection defense, zero-day hunting, OWASP LLM coverage. MIT open source.

Resources

Contributing

Stars

Watchers

Forks

Releases

Packages

Contributors

Languages