ForkFree

Prompts / Agents and automation

Make the AI Finish Every Task

You get every item on a list finished properly, instead of a few done and the rest left as suggestions. It works because it sets the depth of thinking up front, defines "done" as real deliverables, and makes the model count its work before replying.

Worked in our testAll four items are finished in a numbered list, with the count stated up front and the required closing line. Two flaws: the returns-policy note tells the user to adjust placeholders, which breaks the "no next steps" rule, and its stated word count of 72 is actually 73 (still under 80).

The prompt

You are a diligent operations assistant who completes every job in full before replying.

Task: Work through every item in {task_list} for {business_context}, using the depth of thinking set in {effort_level}. Do not stop part-way or hand back a to-do list.

Steps:
1. Read the whole list first and state how many items it contains.
2. Match your depth to the effort level: keep "light" quick and direct, and for "deep" reason carefully and check your work.
3. Complete each item with the real deliverable, such as the finished text or answer, never a plan or a promise to do it later.
4. Before replying, compare the number of items you finished with the number you counted. If any are missing, finish them now.

Output: A numbered list matching the original items, each with its finished deliverable. End with one line reading "Completed X of Y items." Include no next steps and no unfinished work.

Change the parts in curly braces. Use it for long, multi-part jobs such as batches of emails, product copy or data clean-ups, where models tend to stop early.

What happened when we ran it

We sent

You are a diligent operations assistant who completes every job in full before replying.

Task: Work through every item in Write a two-sentence product description for our oat-milk hot chocolate sachets. Draft a polite reply to a customer whose parcel arrived two days late. Suggest three subject lines for our January newsletter about winter warmers. Rewrite our returns policy so it is easy to understand in under 80 words. for a small online café supplier in Leeds that sells coffee, hot chocolate and tea to home customers and has a friendly, plain-spoken brand voice, using the depth of thinking set in deep. Do not stop part-way or hand back a to-do list.

Steps:
1. Read the whole list first and state how many items it contains.
2. Match your depth to the effort level: keep "light" quick and direct, and for "deep" reason carefully and check your work.
3. Complete each item with the real deliverable, such as the finished text or answer, never a plan or a promise to do it later.
4. Before replying, compare the number of items you finished with the number you counted. If any are missing, finish them now.

Output: A numbered list matching the original items, each with its finished deliverable. End with one line reading "Completed X of Y items." Include no next steps and no unfinished work.

The model replied

There are 4 items in this list. 1. **Product description (oat-milk hot chocolate sachets)** Rich, velvety cocoa blended with creamy oat milk, our hot chocolate sachets turn a cold afternoon into something properly cosy. Just add hot water, stir, and enjoy a dairy-free treat that tastes indulgent without any fuss. 2. **Reply to a customer whose parcel arrived two days late** Subject: Your recent order Hi [Name], Thanks for getting in touch, and I'm sorry your parcel reached you two days later than we promised. I know it's frustrating to wait longer than expected, especially when you're looking forward to your order. We've passed your feedback on to our courier so we can try to stop it happening again. As a thank-you for your patience, we'd like to offer you [10% off / free delivery] on your next order. Just use the code [CODE] at checkout. If there's anything else we can help with, just reply to this email and we'll sort it. Kind regards, [Your name] [Café name], Leeds 3. **Three subject lines for the January newsletter (winter warmers)** - Warm up your January: our favourite winter sips - Cold outside? Cosy in a cup, sorted - Brews, blends and hot chocolate to beat the January chill 4. **Returns policy rewritten (under 80 words)** Not happy with your order? Tell us within 14 days of delivery and we'll make it right. Unopened items can be returned for a full refund. Please post them back to us in their original packaging. If something arrives damaged or wrong, email us a photo and we'll send a replacement or refund you straight away. Refunds go back to your original payment method within 5 working days of us receiving your return. *(Word count: 72. Note: the 14-day window, 5-working-day refund time and unopened-item rule are placeholders, so please adjust them to match your actual policy, since I didn't have the original text.)* Completed 4 of 4 items.

Run on Claude Sonnet 5.5, 12 Nov 2026. Automatic check: 4 out of 5.

Mentioned for

Claude

Where it came from

Found on Web, shared by Moritz Kremb (@moritzkremb): Moritz Kremb on X: Anthropic just released a prompting guide for Claude Fable 5.2. Here are the 10 rules.

This is our own version of the idea, written and tested by us. The original may differ.

More agents and automation prompts

Agents and automation Partly worked
Rewrite Agent Rules With Explicit Exceptions

You get a cleaner set of agent instructions where risky actions have a clear approval gate instead of a vague ban, plus progress updates and a cap on helpers. It works because models tend to treat "never" as absolute, so spelling out the exception avoids both blocked work and unsafe guesses.

Found on Web, shared by Anthropic (official docs)
Agents and automation Worked in our test
Audit Subagents Before Folding Them

You get a verdict for each subagent, backed by evidence and a test, plus a safe order for making changes. It works because it forces a measured decision per subagent instead of a blanket rewrite.

Found on GitHub, shared by anthropics
Agents and automation Partly worked
Find Silent Errors in Agent Traces

You get a structured list of hidden mistakes in an agent run, even when every tool call succeeded. It works because the AI checks each step against the stated goal instead of only the final answer.

Found on Hacker News, shared by skull8888888