0.72.0

Changed — may break callers

• memo --help and a bare memo show a short first screen instead of every command: what memo is for, eight
  things it does, the other WireCat tools, and where to go next (memo commands, memo guide, memo help options,
  memo skill show). A script that read a command or an option from the root help reads memo --help-all, which
  prints the old help, or memo help options.
• A command's --help ends with Returns, Next usually and Example in place of its Examples block, worked out
  from the command's definition. docs/commands.md shows the same lines.
• A list command refuses an unknown --option before it reads anything, with commander's message
  (unknown option '--project'). memo tasks list --project HRB used to answer with a pointer to project:HRB.

Added

• memo guide [name]: scenarios, the commands for a common task in order and why each, with --json for agents.
  person-brief gathers one person across tg, MAX, mail, notes, meetings and Drive before a call;
  decision-trail finds what was decided about a topic across every source; start-the-day, prepare-meeting and
  set-up stay within memo. Every memo step is checked against the command tree in CI.

Security

• MEMO_READ_ONLY=1 refuses every change with permission_error (exit 5) before the command runs: writes to
  the store and your files, mail trash and restore, the timer, the skill files, upgrade and ask --allow-remote.
  Reads and memo sync still run, and so do tasks plan and tasks link, which cli-core marks as safe for a
  scheduled agent. memo mcp leaves its writing tools out under the lock.