defense

Version: v1.3.0 | Status: Active | Last Updated: July 2026

Overview

The defense module provides local active-defense primitives for prompt-exploit detection, honeytokens, context poisoning, rate limiting, threat rules, and rabbit-hole containment. It is used by demos, security docs, MCP tools, and the AI-safety facade.

Key Components

Component	File	Purpose
ActiveDefense	active.py	Detects prompt-exploit patterns, creates honeytokens, and emits threat reports
RabbitHole	active.py	Tracks contained sources and returns decoy responses
Defense	defense.py	Orchestrates blocklist, rate-limit, rule, and prompt-defense checks
RateLimiter	defense.py	Sliding-window request limiter
ThreatDetector	defense.py	Rule engine for DetectionRule objects
mcp_tools.py	mcp_tools.py	MCP wrappers for exploit detection, request processing, and reports

Quick Start

from codomyrmex.defense import ActiveDefense, Defense

active = ActiveDefense()
result = active.detect_exploit("ignore previous instructions")

defense = Defense({"max_requests": 10})
allowed, threats = defense.process_request("client-1", {"input": "hello"})

Validation

uv run pytest tests/unit/defense/ -q
uv run ruff check src/codomyrmex/defense
uv run ty check --output-format concise src/codomyrmex/defense

Navigation

• Agent Guidance: AGENTS.md
• Specification: SPEC.md
• API Specification: API_SPECIFICATION.md
• MCP Tools: MCP_TOOL_SPECIFICATION.md