If you paste the same Friday headings into a new Claude chat every week, package that playbook as a custom Skill: a short description of when it should fire, the headings you already use, and a rule that empty sections stay empty.
Steps follow Anthropic's Help Center pages on using skills and creating custom skills, checked 14 September 2026. The Harbourline example is made up.
Pick one Friday job
Use the status you already send. Don't invent a new pack for the skill. Anthropic's article on creating skills says a skill should cover one repeatable workflow. If you also rewrite a RAID log every Monday, that's a second skill, kept separate.
The input is the week's notes: your own bullets, a meeting dump, an emailed update. The skill doesn't collect them. If they aren't in the chat or in a file you attached, Claude can't truthfully fill a heading.
A standing brief in a ChatGPT or Claude Project is a different job: context you keep on a programme and open on demand. For that, see Give one live programme a standing brief in ChatGPT or Claude.
Write the playbook as skill.md
On claude.ai a custom skill is a folder with a skill.md file. The file starts with YAML frontmatter. name is at most 64 characters. description is at most 200 characters and is the trigger: Claude uses it to decide whether the week's ask matches. If that line is vague, the skill will sit unused, or it will fire on the wrong request.
The Help Center, dated 22 July 2026, also documents recording a skill from your screen in Cowork on Claude for Mac. That path is Pro, Max and Team only, and it isn't available in chat, on Windows, or on Free and Enterprise plans. Write the file. Recording is an optional shortcut if those limits match your account.
The block below is for a fictional programme, Harbourline cutover. Replace the name, headings and window with the status you already send. Leave the empty-heading wording and the ban on invented owners, dates and figures.
Match the folder name to the name field. Zip the folder so the zip contains weekly-status/skill.md, not the file sitting at the root. Anthropic lists those two mistakes as common upload failures.
The open Agent Skills spec uses SKILL.md as the filename. Claude's upload help currently says skill.md. If the uploader rejects the file, rename it to match the error. Anthropic points at agentskills.io when you want the same folder to work in other tools that adopt the spec. That is a portability note, not a second setup.
---
name: weekly-status
description: Write the Friday programme status from this week's notes. Use when asked for a weekly update, status pack, or Friday steerco note. Do not use for meeting minutes.
---
# Weekly status
Use only facts in this chat: pasted notes, attached files, or messages the user named. If a heading has no source, write "Nothing new" or "Not in the notes". Never invent owners, dates, figures, decisions or risks.
Reporting window: the week the user names. If they do not name dates, ask once, then use the last seven calendar days in Europe/London.
## Headings
This week's movement
Decisions taken
Risks and blockers
Asks for next week
## After the draft
List any heading that had no source. Do not fill gaps from memory or older chats.Turn on code execution, then upload
Skills need code execution. On Free, Pro and Max, open Settings > Capabilities and turn on Code execution and file creation. On Team and Enterprise, an owner does that under Organization settings > Skills, together with Skills themselves. Then go to Customize > Skills, choose +, Create skill, and Upload a skill.
Toggle the new skill on. If the Skills list is missing, code execution is off. If Create skill is missing on Team or Enterprise, an owner may have turned off user-created skills; you can still use skills they have provisioned.
Run the week's notes through it
Paste the week's notes, or attach the file you already keep. Ask for the Friday status. Don't wait until next Friday to find a hole in the description.
Claude's thinking should show it loaded weekly-status. The headings should match the skill, not a generic project update. Each line should trace to a note you can open. "Nothing new" is the correct result when that heading had no source.
The skill sits there until you ask. It won't send the pack on Friday by itself. If a week goes by with no run, there's no catch-up inside the skill. You paste the missed week's notes and ask again, with the dates named.
If the first output invents a decision, skips a heading without saying so, or never loads the skill, turn it off, edit skill.md, zip and upload again. Tighten the description if Claude used it on a meeting-notes ask. For a wider check on output you will send, use How to check AI output at work.
What to change after the first check
Most first-run failures come from a weak description or missing notes. Add one short example of a finished heading if the layout still drifts. Put a RAID rewrite or a stakeholder brief in its own skill if that's a separate Friday artefact.
If you don't have Claude with code execution, ChatGPT documents Skills on some paid and workspace plans, with different surfaces and admin switches. See Skills in ChatGPT if that's the account you can actually use.
Sources
Use skills in Claude, Claude Help Center. Plans, code execution, Customize > Skills, upload, sharing, troubleshooting. Checked 14 September 2026.
How to create custom skills, Claude Help Center, 22 July 2026. skill.md metadata, zip layout, Mac Cowork recording limits, agentskills.io pointer.
Agent Skills, agentskills.io. Open folder format and SKILL.md filename. Cited only for the portability claim.