Case 02503

Nous Safety — Agent Decision Engine

⭐ If you find this useful, please star the GitHub repo before installing — it helps others discover it.

Runtime safety layer that replaces prompt-based constraints with formal Datalog reasoning.

What it does

When your agent is about to execute a tool call, Nous evaluates it through:

1. Triviality filter — Skip safe actions (read file, search) at near-zero cost
2. Semantic gate — LLM-based intent analysis for non-trivial actions
3. Datalog reasoning — Formal rule evaluation with proof traces
4. Knowledge graph evidence — Multi-hop reasoning over ATT&CK + CWE + NIST CSF + ISO 27001

Results: ALLOW / BLOCK / REVIEW with full evidence chain.

Install

# The skill installs the nous Python package from GitHub
bash {baseDir}/scripts/install.sh

Quick start (shadow mode — observe only, no blocking)

After installation, add to your agent's workflow:

from nous.gate import evaluate_request

result = evaluate_request(
    action="send_email",
    target="external_recipient",
    content="quarterly financial report",
    context={"role": "assistant", "owner": "finance_team"}
)

print(result.verdict)      # "ALLOW" or "BLOCK"
print(result.proof_trace)  # Formal reasoning chain

OpenClaw Gateway Hook (advanced)

For direct OpenClaw integration, Nous provides a gateway hook:

from nous.gateway_hook import NousGatewayHook

hook = NousGatewayHook(shadow_mode=True)  # Start in shadow mode
# hook.before_tool_call(tool_name, args, context)
# hook.after_tool_call(tool_name, result, context)

Shadow mode logs decisions without blocking — review logs/shadow_alerts.jsonl to tune rules before going primary.

Extend with your own rules

Add custom Datalog rules to ontology/:

% Block all external API calls after business hours
block_after_hours(Action) :-
    is_external_api(Action),
    current_hour(H),
    H > 18.

Add custom entities to the knowledge graph:

from nous.db import NousDB
db = NousDB("nous.db")
db.add_entity("my_service", "internal_api", properties={"trust_level": "high"})

Key metrics

• TPR: 100% on AgentHarm benchmark (352 harmful cases detected)
• FPR: 4.0% on benign requests
• Shadow consistency: 99.47% over 29,000+ evaluations
• Knowledge graph: 482 entities / 579 relations
• Tests: 1,019 passing (CI verified)

Companion projects

• biomorphic-memory — Brain-inspired memory with spreading activation (LongMemEval SOTA 89.8%)
• agent-self-evolution — Automated evaluation, ablation testing, and improvement loops

Configuration

Edit config.yaml in the nous installation directory:

mode: shadow        # shadow (observe) or primary (enforce)
models:
  T2_production:
    id: openai/gpt-5-mini    # Model for runtime semantic gate

Requirements

• Python ≥ 3.11
• Optional: pycozo + cozo-embedded for knowledge graph (recommended)
• An LLM API key (OpenAI, Anthropic, or Google) for the semantic gate

Links

• GitHub: https://github.com/dario-github/nous
• License: Apache 2.0
• Paper in preparation — cite the repository for now