XERXES SI · SECURITY DECISION LATTICE
You have a question. Follow the boundary.
Don't guess what your agent can reach. Choose your situation, see the underlying risk, inspect the control, and follow the evidence. Every path remains visible to crawlers and works without JavaScript.
Find the question you actually need answered
Start typing to filter short answers and evidence links. All primary paths are also available below.
I’m installing at homeStart with local files, browser credentials, and how much access the new assistant really needs.Get the evidence →I have important private dataCheck mounts, credentials, LAN routes and read-only tasks before introducing broad agent tools.Get the evidence →I want a security architectureUnderstand why independent enforcement and approval boundaries matter.Get the evidence →I’m managing several agentsEvaluate budgets, enrollment, approvals, maintenance and restoration.Get the evidence →I’m comparing productsCredit native capabilities and compare the complete operating workflow.Get the evidence →I’m evaluating or funding a platformRequire proof of denial, reproducible deployments, economics and operator accountability.Get the evidence →
From question to accountable control
- Situation: what am I asking the agent to do?
- Exposure: which data, tools, network services or credentials are reachable?
- Enforcement: which sandbox, policy, identity and approval gate actually constrains it?
- Operation: who handles resources, updates, review and recovery?
- Evidence: which logs, denials and host checks demonstrate the policy?
See XMAN engineering evidence and the independent pilot checklist. The goal is to verify protection, not declare any single product universally safe.