52 Hours a Year: A Claude Code Skill for Weekly Reporting
An hour a week on reports is 52 hours over 52 weeks, so $5,200 at $100 an hour. A lot of jobs want a weekly report of some kind.
⏱️ The cost: minutes a week x 52 x your hourly rate.
🧩 The skill: pulls the week's numbers and drafts the report. You check it.
💵 The build: $250 once. It pays back when the time saved passes $250.
What does a Claude Code skill for weekly reporting do?
The SKILL.md doesn't need much in it, mostly where the numbers come from and a copy of last week's report.
↓
📂 2. Claude reads the skill file from your own disk
↓
📥 3. A script pulls the week's numbers: an export, a CSV, an API
↓
✍️ 4. Claude writes the summary in your format
↓
📄 5. The draft lands in a file for you
↓
✅ 6. You read it, fix anything off, then send it
What is your weekly report worth over 52 weeks?
I worked these out at $50, $100 and $150 an hour since I don't know what anyone's time is worth.
Work out your own number
You'll probably have to guess the minutes, I did. Checking the draft takes a while too so I put 15 in for that.
Saving 45 minutes a week is 39 hours over 52 weeks, worth $3,900 at $100 an hour. A $250 build pays back in 4 weeks.
The 60 and 15 minutes are examples too. If checking takes nearly as long as writing it, the skill isn't worth building.
The weekly report I've got
The only weekly report I have at the moment is a script that counts the views on each post on this blog. It has run once.
A dishwasher is probably the closest thing I can think of, you still load it but you don't do the washing.
Can it run without you?
Yes, claude -p runs Claude Code without the chat window, and you can put /weekly-report in the prompt. Point a scheduled task at that and the draft's there on Monday.
This is the line the scheduled task would run:
claude -p "/weekly-report" --allowedTools "Read,Write,Bash(python *)"
I'd still read it before it goes to a client. There's more on leaving a chore running in whether Claude Code can run a chore without you watching.
When it's worth building
It works when the numbers come from the same places every week. If half the report is your take on how the week went, it can only draft the other half.
Good fit
- Same sources every week
- Same layout every week
- Numbers sit in an export, a CSV or an API
Poor fit
- The numbers live in someone's head
- The format changes every week
- It's mostly your opinion
What does one cost?
You can write it yourself for free with our step-by-step skill guide. For a rough feel first, see how many hours a Claude Code skill saves.
Or I build it for a fixed $250, and you agree in writing what it does before I start: get a custom Claude Code skill built.
- A skill is a SKILL.md folder, can run scripts, starts with /name: Claude Code docs, code.claude.com/docs/en/skills ("Skills can bundle and run scripts in any language"; "by typing / followed by the skill name"), fetched 2026-10-05.
- claude -p runs it without the chat window, /skill-name works in the prompt: code.claude.com/docs/en/headless ("Add the -p (or --print) flag to any claude command to run it non-interactively"; "Include /skill-name in the prompt string and Claude Code expands it before running"), fetched 2026-10-05.
- The view-count script has run once: this blog's own weekly traffic table, one week so far, to 19 September 2026.
- $250 fixed: the custom build page.
- Every other figure: arithmetic on example numbers you can swap for your own, not survey data.
Related: a Claude Code skill for bookkeeping reconciliation and a Claude Code skill for invoice chasing.
Get the hour back
Send the report you write every week. You get a written scope first, then a fixed $250 build that runs on your own plan.
Get a written scope