What it does
You paste in the week’s raw material: your notes, workstream leads’ updates, last week’s report. It gives back a draft weekly status in RAG format: an overall status, the asks, then one row per workstream with a rating in words, a one-line reason and an ask with an owner.
It drafts. You still decide every rating. The value is the forty minutes of sorting and rewording it takes off your Friday, not the judgment.
When to use it
| Use it when | Skip it when |
|---|---|
| You have notes from several workstreams and need one report | You haven’t talked to the workstream leads this week: it will confidently summarize stale notes |
| The report format is settled and the same every week | The program is in crisis and the report is really a decision memo |
| Your notes are safe to paste into the AI tool your company allows | The notes hold anything your AI policy says stays out |
The input
Give it three things, pasted as plain text under the prompt:
- Last week’s report, so it can say what changed.
- This week’s notes: your meeting notes, workstream leads’ updates, chat threads worth keeping. Messy is fine.
- Your RAG definitions, if the team agreed on them at kickoff. Without them it uses the ones in the prompt.
The prompt
You are helping a technical program manager draft this week's program status report.
Use only the material I paste below. Do not invent dates, names, numbers or progress. If something a report needs is missing, write [MISSING: what is needed] in its place.
Rate each workstream:
- Green: on track for its next milestone date.
- Amber: at risk; the date holds only if a named problem is fixed soon.
- Red: the date will slip without a decision or help from outside the team.
Rate against the date, not the effort. When the notes are unclear, choose the less optimistic rating and say why.
Write the report in this order:
1. Overall status: one rating and one sentence on why.
2. Asks: each ask names who it is for and the date it is needed by.
3. A table with columns Workstream, Status, Why, Ask. Write the status as a word, not a color or emoji. Every Amber or Red row needs an ask with an owner.
4. What changed since last week: at most three bullets.
5. Questions for me: anything in the notes that contradicts itself or that you could not rate.
Keep it under 300 words, excluding the table. Plain language, no filler.
LAST WEEK'S REPORT:
[paste]
THIS WEEK'S NOTES:
[paste]
OUR RAG DEFINITIONS (optional):
[paste]Setup notes. In a chat assistant, paste the prompt and your material in one message. In a coding agent or CLI, save the prompt to a file and point the agent at your notes file. With a project or custom-instruction feature, save everything above LAST WEEK'S REPORT as the instructions so each week you only paste the material.
A filled example
From the running example, Lantern: five teams moving a support assistant onto in-house inference. These are the notes the TPM pasted in.
| Source | What it says |
|---|---|
| Data lead, Tuesday | Index rebuild running about half the planned speed. Changed the batch size on Nov 12. Too early to say if it helped. |
| Cutover sync | Shadow run 1 done, two runbook steps out of order, fixed. Shadow run 2 next week. |
| Security, chat thread | Still waiting on a reviewer for the security & privacy review. Nobody owns scheduling it. Review gates customer traffic. |
| Plan review | All five teams signed off on the migration plan Monday. |
| Last week’s report | All four workstreams Green. |
And the draft it gave back:
| Workstream | Status | Why | Ask |
|---|---|---|---|
| Migration plan | Green | Signed off by all five teams on Monday | None |
| Index rebuild | Amber | Rebuild at about half the planned speed; the Nov 12 batch-size change may fix it | Data lead to confirm the fix worked by [MISSING: date] |
| Cutover | Green | Shadow run 1 complete; two out-of-order runbook steps fixed | None |
| Security & privacy review | Red | No reviewer assigned and no owner for scheduling; the review gates customer traffic | Security lead to name a reviewer by [MISSING: date] |
Above the table it wrote Overall: Amber, because the security & privacy review gates customer traffic and has no owner, and asked one question: did the Nov 12 batch-size change speed the rebuild up? That question is the point. The notes didn’t say, and a confident Green would have been wrong.
What to verify
| Check | Item |
|---|---|
| Every rating matches what you heardOverrule any rating that surprises you; you were in the meetings, the model wasn’t. | |
| No [MISSING] markers leftFill each one, or delete the claim that needed it. | |
| Every ask has a person and a dateAn ask without an owner is a wish. | |
| Nothing was inventedCheck each date, number and name against your notes. | |
| The overall status is not the averageOne Red that gates launch makes the program at least Amber. | |
| Questions answeredRead the “Questions for me” section; it often finds the contradiction you missed. |
When not to use it
- When the status is the news. If a workstream just went Red and you need a decision this week, write that message yourself. A drafted status table softens it.
- When you haven’t done the legwork. The prompt summarizes what you give it. Notes you didn’t gather are risks it can’t see.
- When your notes can’t leave your company’s approved tools. Check your AI policy first, and strip out anything it keeps in.