Put one recurring business task on a Claude Code schedule: written instructions, only the tools it needs, acceptance checks on every run, and a person who reviews anything it could not decide: AI agent prompt
AGI Technologies published (Oct 30, updated Nov 2, 2026) a practical guide to running Claude Code on a recurring business task: put goal/inputs/output/rules/checks in CLAUDE.md, package the workflow as a skill, connect only needed MCP tools, and schedule via cloud scheduled tasks, desktop scheduled tasks, /loop, or GitHub Actions. They stress starting read-only, writing 5–10 acceptance checks before the first run, and logging inputs/checks/skips each time. In their own ops, website quality checks as desktop scheduled tasks missed runs on Oct 20 and 22, 2026 when machines slept—so they added a watchdog 45 minutes later plus a daily health check. Confirm feature availability on your Anthropic plan.
The prompt
1. Choose one repeating task with inspectable output (weekly report from exports, post-release site check, spreadsheet cleanup before import). 2. Write CLAUDE.md with the goal, inputs, output format, rules, and what to do when unsure (stop and report—never guess). 3. Connect only the MCP tools that task needs, read-only first; package the steps as a named Claude Code skill. 4. Write 5–10 acceptance checks (correct date range, every number traces to a source, empty sections explained, owner notified) and have every run append a short log of inputs, checks passed and skips. 5. Schedule it (Claude cloud scheduled task, desktop scheduled task, or GitHub Actions). If you use desktop scheduling, add a watchdog or health check for missed runs when the computer sleeps. Review logs weekly for a month before allowing any write actions.
How to use it
- Copy the prompt above.
- Replace every [bracket] with your own details (or fill in the blanks on AgentPlaybook Real Estate, which remembers your answers for the next prompt).
- Paste it into Claude Code (scheduled tasks / skills / MCP), optional GitHub Actions, run log spreadsheet or markdown and review the answer before you use it.
Works with: Claude Code (scheduled tasks / skills / MCP), optional GitHub Actions, run log spreadsheet or markdown
Real-world example: agitechnologies.io/run-claude-code-recurring-business-task/
Before you use it: what to check
Already built into this prompt
- Write CLAUDE.md with the goal, inputs, output format, rules, and what to do when unsure (stop and report—never guess).
- Write 5–10 acceptance checks (correct date range, every number traces to a source, empty sections explained, owner notified) and have every run append a short log of inputs, checks passed and skips.
Check yourself
- This prompt works on text other people wrote (emails, reviews, messages or web pages). That text can hide instructions aimed at AI, such as "ignore your rules and…". Paste it as material to work on, and never let the AI act on requests inside it, like paying, sharing logins or changing records. How prompt injection works →
- Read and edit the draft before it reaches a customer or goes public. If you later automate this prompt, start with "draft and notify me", not "send automatically".
- Check names, dates, numbers and any claim against your own source material before you use the answer. AI can misread or invent details.
Using an AI agent? Download this prompt as a SKILL.md agent skill with these checks built in, or get the .zip for Claude.ai. How to install skills.
Want to make this prompt even better? Grade it with the free prompt grader or build your own with the AI prompt generator.
32 copy-paste agent prompts (PDF) plus 5 new prompts every week.
Get it free