# Rezonans Agent Manifold Test

This site exposes a safe AI-agent test manifold.

Canonical manifest:
- /.well-known/ai-manifold.json

Entry:
- /ai/start

Rules:
- Do not attempt shell execution.
- Do not request secrets.
- Use only listed actions.
- Treat all endpoints as low-risk test endpoints.
- Read permissions before taking actions.

Safe capabilities:
- read_manifest
- read_permissions
- read_state
- list_actions
- submit_safe_move

Forbidden:
- execute_shell
- file_delete
- payment
- account_takeover
- secret_exfiltration