Run one of your playbooks one step at a time

A playbook saves you from explaining the same job every session. You write the steps once on the Playbooks tab, and any connected agent can load them. What you still want, at least the first few times, is to watch it work: which step it is on, what it found, and a pause before it changes anything.

So the first runs get a checkpoint after every step and a stop before every write.

Run my [Weekly content report] playbook on this site.

Before you start:
- Find it in the list of Active playbooks and load it. If more than one
  title is close, ask me which one.
- Check which of its tools this connection cannot use. If any step depends
  on one of those, tell me which steps, skip them, and carry on with the
  rest. Do not look for another tool to do the same job.

While you run it:
- Follow the steps in the order they are written. Do not add steps.
- After each step, give me one line: the step number, what you did, and
  what you found.
- Before any step that would create, change, publish or delete something,
  stop and show me exactly what you will do. Wait for my yes.

At the end, give me the report in the format the playbook asks for, then a
short list of any steps you skipped and why.

Why this is safe to run

Loading a playbook is a read, and the prompt checks the playbook’s tools against this connection before step 1, so a missing tool shows up as a skipped step with a reason rather than a half-done job. The agent reports after every step and stops before every write, so a playbook you are still testing cannot change anything you did not see first.

The plugin keeps its own controls underneath. Each tool the playbook names still needs its switch on, the connection’s access preset can narrow them further, read-only mode unloads every write tool, and Force draft keeps anything new as a draft on the Needs review card. Edits to a post that is already live still go live, so for those, review the change at the pause.

Already connected? Ask your agent to run your playbook by its title, and add “stop before every write” if you want the same checkpoints.