Reflect and plan tomorrow with the PM agent

The hardest part of solo or near-solo work is not the doing, it is the deciding. Every morning you have a hundred things you could do. The wrong choice is invisible. You only realize on Friday that you spent the week on the wrong work.

The PM agent solves this. Not by being smarter than you. By holding the context across days: reading the product plan, the QA reviews and yesterday's commits, knowing what is open, and putting choices in front of you in the right order. You make the calls; the PM agent does the staging.

The hardest part of solo work is not the doing, it is the deciding

Every morning you have a hundred things you could do. The wrong choice is invisible. You only realize on Friday that you spent the week on the wrong work.

The PM agent solves this. Not by being smarter than you. By holding the context across days, reading the status file, knowing what is open, and putting choices in front of you in the right order. You make the calls; the PM agent does the staging.

The reflection, at the end of the day

Ten minutes before you close the laptop. It is the single habit that makes tomorrow start fast.

• The PM agent reads the day
  The product plan, every review in docs/qa, today's commits, anything still marked pending. It writes what is live (URLs), what passed QA and what did not.
• It asks you three questions
  What slowed you down most today? What surprised you? What is the one thing you want working by the end of tomorrow? Answer honestly. Nobody reads this but you and your agent.
• It leaves notes for tomorrow
  What to look at first. What is still broken. What today taught you about the next build. Written to Working Files/day-N-reflection.md, committed and pushed, so it is there on any machine in the morning.

The plan, the next morning

Same shape every day. Twenty minutes. Sets the rest of the day.

• Open Claude. Greet the PM agent
  "Good morning. What did we ship yesterday and what should we do today?" That sentence runs the entire flow.
• PM reads context
  It opens yesterday's reflection, the product plan and the open epics, reads yesterday's commits, and checks your calendar if connected.
• PM proposes a plan
  Three to five things, in priority order, each tied to an open epic. Each one explicit about which agent (or you) does it.
• You approve, edit, or replan
  You are still the decision-maker. Replan freely. The PM does not get its feelings hurt.
• PM updates the plan file
  Writes docs/plans/[today].md. That file becomes the day's contract. Anyone (or any agent) who jumps in mid-day reads it and knows where things stand.

Subtleties most people miss

Three things that make this work over weeks, not just one morning.

• The PM does not assign work to itself
  It plans the work, you do it (or another agent does it). The PM that does its own work loses the planning edge fast.
• The PM reads what shipped yesterday
  Yesterday's commits, yesterday's reflection, yesterday's QA review. That continuity is what makes the PM feel like a partner instead of a daily reset.
• The PM flags drift
  If you have been In Progress on the same epic for nine days, the PM notices and asks. Drift is invisible to you; you live in it. The PM has fresh eyes every morning.

What this is not

Not a calendar (your calendar stays the calendar). Not a productivity coach (no nudges, no streaks). Not a journal (no feelings logged). It is a planning partner. One specific job, done well.

A reflection on a real day and a plan for the next one, both written by the PM agent and saved to disk. The ritual practiced once. The plan survives a restart, a different chat, and a different machine.