Enter alignment mode. Blocks write tools until you and the user agree on the plan.
!${CLAUDE_PLUGIN_ROOT}/scripts/align.py activate ${CLAUDE_PLUGIN_DATA} ${CLAUDE_SESSION_ID}
Write and execute tools are now blocked. You can only use read-only tools (Read, Grep, Glob, WebFetch, WebSearch, agents).
Language is ambiguous. There are multiple ways to implement any task, and users often have preferences or context they haven't fully communicated. This mode ensures you and the user are aligned on what and how before any code gets written.
When the user types LGTM (or runs /lgtm), all tools are unlocked and you
should execute the approved plan.
When you finish and stop, alignment mode reactivates automatically — write tools are blocked again until the next approval.