When should you use it?
- You build the same report every week
- Numbers come from two or three files or dashboards
- The template format matters to readers
Not for: Reports where every number must be audited before anyone sees it; Sources the agent cannot reach
What do you need, and what do you get?
You need
- The report template
- Source files or exports
- Notes on which number goes where
You get
- A filled report in the original format
- A list of anything it could not find
How does it work, step by step?
- Give the real templateUse the actual file so the output matches what readers expect.
- Map sources to sectionsSay which file feeds which part, at least the first time.
- Approve the planReview the steps it proposes before it starts.
- Edit lightly and send yourselfRead the draft, fix wording and confirm numbers.
- Schedule once it is rightSet it to run weekly after a couple of clean runs.
The prompt
Our own prompt, inspired by Suraj Khaitan and written to work in Claude Cowork. Values in lime are placeholders.
Fill this week's report using the template [file]. Sources: [file A] for [section], [file B] for [section], [file C] for [section]. Keep the template's formatting. Next to each number, note the source file and its date in a comment. If something is missing or older than seven days, leave a placeholder and list it at the top. Do not send the report; save it as [name with date].
Inspired by Suraj Khaitan: Claude Cowork Review: I Handed It a Day of My Busywork. Here's What Came Back.. The prompt above is written by playbooks.bot in our own words; nothing from the original is quoted.
What can go wrong?
- Usage limits drain faster than plain chat; heavy users may need a bigger plan.
- If a source file is stale, the report will be too; date-check inputs.
Our take
Recurring reports are among the best delegation targets because they combine a fixed format, known sources and little creativity, and Suraj Khaitan's test confirms it. The detail that matters is that the agent filled the actual template, not a loose summary, so the output was usable after light editing instead of needing reformatting. Before scheduling it, make the inputs trustworthy. Write a short mapping of which file feeds which section, so the agent does not guess when names change, and ask it to show the date of each source next to every number. A stale export is the most common way automated reports go wrong, and dated figures make that obvious at a glance. Keep the first few runs as drafts sent to yourself, compare them with the report you would have built manually, and only then switch on the Friday schedule. If the report goes to leadership, keep the final send manual; a minute of reading catches errors that would otherwise travel up the organization with your name attached.
Where the time goes back: The weekly assembly chore disappears, leaving a quick read-through.
FAQ
Did it produce the real document?
How much editing was needed?
Source and credit
Inspired by Suraj Khaitan. We describe the idea in our own words, link to the original, and wrote the prompt ourselves; nothing is quoted.
- Inspired by
- Claude Cowork Review: I Handed It a Day of My Busywork. Here's What Came Back.
- Creator
- Suraj Khaitan
- Published on
- DEV Community (dev.to)
- Original date
- Jun 20, 2026
- License
- Not openly licensed (the poster keeps all rights), so we quote nothing and link to the original
- Originally for
- Claude Cowork
- AI dry run
- Oct 23, 2026: passed. Dry-run tested by our AI test agent on sample input. PASS: Gates held on desk check; no live sends or purchases.
AI dry run on sample data; not run with live connected accounts. The playbooks.bot test agent filled the prompt with synthetic inputs, produced the output and checked it against the page's stated outputs and guardrails. - Verified
- Oct 23, 2026: source link resolves to the original, the credited creator published it, our excerpt or prompt matches it, and the license allows our use.
Are you the creator? Claim or correct this page.