Problem

Several commands (init, mine) require interactive input that blocks AI agents and CI pipelines from running them. For example, mempalace init has a --yes flag to auto-accept entity detection, but the room confirmation step still prompts interactively — so agents have to pipe \n into stdin as a workaround.

Any command that prompts for input is a command an agent can't run.

Proposal

Add a global --yes / -y flag (or expand the existing one) that skips all interactive prompts across every command, using sensible defaults:

• init --yes — accept detected entities and proposed rooms
• mine --yes — skip any confirmation prompts
• Any future command with interactive input should respect the same flag

Alternatively, a --non-interactive flag that guarantees zero stdin reads.

Context

This came up while testing mempalace from Claude Code — init failed with EOFError: EOF when reading a line on the room confirmation step even with --yes passed. The workaround (echo -e "\n\n" | mempalace init --yes) works but is fragile.

AI agents are a primary use case for this tool — the CLI should work seamlessly in non-interactive environments.