scope-enforcement

27 scope-enforced AI agents that run the full pentest kill-chain (recon → exploit → post-ex → DFIR → report) as a one-command Claude Code plugin. Backed by 754 MITRE-mapped skills.

A Claude Code skill that keeps edits surgical. Detects out-of-scope modifications and flags unasked changes before they happen.

Specification and architecture preview for authorized, evidence-backed offensive operations.

MCP server that enforces multi-agent scope boundaries — prevents scope drift, work duplication, and infinite delegation loops in agentic systems

A grounded answer engine for LLM assistants: answer only from your approved data, refuse out-of-scope asks, and prove it with an eval set. Offline-first; OpenAI/Anthropic optional.

🐕 Autonomous pentest framework for Claude Code with enforced, deny-by-default scope safety — recon → bounded exploit → report, one command, fully audited.

Canonical spec + RFCs for Agent Authority Protocol (AAP): human authority, proposal/evidence lifecycle, auditable world-state transitions.