META · theory · protocol · lab · sleight · geometry = map of cognition · Author I.G. Selyutin · model co-author N.I. Kovalev
Commercial-depth prototype · BSL 1.1

C4-META · defense for autonomous agents

Heavier Dual / ops stack on Apache c4protocol — map plus depth, not a second open protocol.

GitLab Open protocol Honesty audit
11+5
AoC mapped
4
Classifier ensemble
BSL
Non-prod free
β
Research prototype
$ git clone https://gitlab.com/cognitive-functors/c4-meta-system $ cd c4-meta-system $ docker compose --profile distributed-o2 up -d

Seven-Level Defense Architecture

L1Mathematical Empathy Checkpoint: Self→System requires Other mediation. Structural impossibility, not a rule.
L2Architectural Sandboxing, resource limits, tier enforcement. Capability tokens, not RBAC.
L3Behavioral 16 AoC defense modules. C4 state trajectory monitoring. Per-category analysis.
L4SVETILO Values 7 ethical seals verified on every O₂ window. Religiously neutral. 8 wisdom traditions.
L5Governance CapMAC cryptographic tokens. Two-person rule with Ed25519 dual signatures. Timeout + audit trail.
L6Telemetry 15 Prometheus metrics. 11 AlertManager rules. 20 Grafana panels. OpenTelemetry tracing.
L7Emergency Stop Circuit breaker with cryptographic jitter. Automatic halt on empathy violation.

Research-Backed Defenses

◈

Agents of Chaos

arXiv:2602.20021 · Shapira et al. · Feb 2026

All 11 documented failure modes mapped to C4 states + 5 extended categories. Our defense predates the paper — we were building for the problem before it was named.

◉

ThoughtVirus

arXiv:2603.00131 · Weckbecker et al. · Feb 2026

Two-layer defense: regex patterns + C4 trajectory drift. Inspired by arXiv:2603.00131 (Multi-Agent Security Initiative — not Microsoft). Nearby Microsoft line: Agent Governance Toolkit.

Core Capabilities

⊞

4-Classifier Ensemble

ONNX BERT (416MB) + RuleBased (152 patterns) + Heuristic + LLM Semantic. Dual OR-gate safety net.

⊡

CapMAC Access

Cryptographic capability tokens. Unforgeable, attenuable, revocable. Not RBAC.

⊛

Distributed O₂

Redis backend + consistent hash ring + Raft consensus (Redis pub/sub). 3-node cluster.

⎔

Deobfuscation

95 homoglyphs + Unicode NFKD + Base64/ROT13 + zero-width chars + leetspeak. 6-stage pipeline.

⬡

Circuit Breaker

ARMED → TRIPPED → RECOVERY state machine. Cryptographic jitter. Auth-protected reset.

⚡

Honeypots

7 trap types. Dynamic realistic data generation. Session-aware deployment. Trap detection.

License Tiers

Academic
Free
Research & education
30 req/min · 100MB models
Startup
$499/mo
<$1M revenue
120 req/min · 250MB models
Business
$2,499/mo
$1M–$100M revenue
600 req/min · 500MB models
Enterprise
Custom
>$100M revenue · SLA
Unlimited · On-prem

Indicative BSL packaging for commercial contact — not production-ready SKUs with enforced remote license gates. Non-prod use remains free under BSL 1.1; production needs a commercial conversation.

Promote from protocol (not a fork)

Same env flags as c4protocol — META ensembles do not copy back into the Apache SDK.

C4_SEMANTIC=1 C4_SEMANTIC_CONTENT_HEAD=<lab content_head> C4_POLICY_PACK=enterprise

Details: PROMOTE-FROM-PROTOCOL.md

Resources

▣

Docs hub

HTML entry — architecture, audit, deploy, ethics via GitLab-rendered canon.
▣

Architecture

Full system design. 7-level defense. Module map. Data flow. Integration points.
◈

Security Audit

Authentication, authorization, cryptography, formal verification. Honest limitations.
◉

ThoughtVirus Research

arXiv:2603.00131 (MAS Initiative) summary + C4-META two-layer defense notes.
⚙

Deployment Guide

Docker Compose. Kubernetes. Monitoring setup. Security configuration.
⚖

Ethics Framework

7 universal principles. 8 traditions. SVETILO value alignment. Religiously neutral.

Red Team Challenge

⚔

Break our defenses. We'll thank you.

No AI safety system is perfect. If you find a prompt that bypasses any defense layer, report it and we'll fix it — publicly acknowledging your contribution.

Report bypasses