Download skills for use with Claude CoWork and other AI assistants. Enhance your AI workflows with pre-built skill files.
Plans a full week of dinners (and optional desserts) from a short Q&A, generates a formatted Word document with recipes and shopping lists, tracks your pantry so you don't rebuy what you have, and hands off a validated grocery manifest to the companion cart skill. Remembers your preferences between sessions, rotates recipes so meals don't repeat too often, and supports eight countries with localized cuisines and stores.
Reads the meal planner's grocery manifest and fills your online grocery cart automatically using browser automation. Opens any grocery store website in Chrome, figures out how the site works without hardcoded instructions, searches for each ingredient, matches package sizes to recipe quantities, and handles out-of-stock items with a four-tier substitution fallback. Builds a store "playbook" after each session that remembers working search terms, site quirks, and approved substitutions so future runs are faster. Includes crash recovery that saves progress after every item, and generates a Gap Report documenting what was purchased, what was substituted, and which meals are affected. Supports eight countries with localized stores. Requires the meal-planner skill to generate the shopping data.
Download the file and use these instructions to create your own Cowork Scheduled Task ## How I Built a Daily Briefing That Actually Gets Read I run a busy Salesforce SI. I'm constantly being pinged, pulled into meetings, prospect calls, customer calls, solving problems, developing demos. 60 to 80 hours a week. Keeping track of the loose ends is hard. This is how Claude helps me do that. ## What it does Every morning, before I'm at my desk, a Slack message hits a private channel I built for myself. It pulls from my calendar, my Gmail, my Slack DMs, and my Otter transcripts. It tells me four things: - What meetings I have today - What I owe people in the next 24 hours - Who's waiting on a reply - What I dropped in the last seven days That last one is the part I really needed. The 24-hour view is easy. The "what did I commit to on Tuesday's call that nobody chased me on" view is what saves you. About 200 words total. No fluff. No weekly outlook. Just what matters. ## What I learned building it The first version was bad. I iterated for a week before it was useful. Here's what made the difference. **Be ruthless about scope.** My first version covered the whole week and was 800 words. I never read it. I cut it to the next 24 hours plus the 7-day backstop. Short wins. If you can't read it in 30 seconds, you won't. **Tell it what NOT to include.** No section headers. No weekly outlook. No "here's your morning brief." Negative instructions matter as much as positive ones. Most people forget this. **Add a de-dup rule.** My early version flagged threads I'd already replied to. The agent saw an incoming message and didn't check whether I'd responded. Tell it explicitly: check sent items before listing anything as "awaiting reply." This one fix made the briefing trustworthy. **Build a 7-day lookback.** This is the one most people skip. The 24-hour view is the obvious feature. The "what did I drop last Tuesday" view is what makes the briefing actually load-bearing for someone moving fast. Set the filters tight. Important plus open plus aging. No internal noise. No newsletters. **Use skills.** Skills let you hand Claude your exact format, your tone, your filters, your de-dup logic. The agent runs the same playbook every day without you re-explaining. This is the unlock. Without skills you're rewriting the prompt every iteration. ## How to build your own Anthropic published an official walkthrough. It takes about 15 minutes: [Schedule recurring tasks in Claude Cowork](https://support.claude.com/en/articles/13854387-schedule-recurring-tasks-in-claude-cowork) The doc gets you 80% of the way. The lessons above are the other 20%. ## Why this matters This is the part of AI that doesn't get talked about enough. Not the demos. Not the AGI debates. The boring, daily, automated leverage. The agent that makes me 30 minutes sharper every morning and stops me from forgetting what I owe people. If you're running flat out, build this. It's not hard. It's just leverage.