Updated November 2026 · Rechecked this month: the six agency pages, the recall figures behind the prompt, and the research on LED light and sleep.
How I built Worry Wart, a Grok Bot agent that reads the government recall pages every day and tells me only about the ones that could put someone in hospital.
The short version: this is episode 2 of a series where I build small agents for ordinary life. An agent is software you give standing instructions to once, then leave running. Worry Wart watches six US safety agencies for serious recalls, and twice a month it looks for everyday products people have started to ask questions about. You do not need to write code. You name an agent and paste in one long message, and the whole message is further down this page.
You half-hear that some product might be bad for you. You notice there was a recall on something but miss which brand. None of it is urgent enough to sit down and research, so it stays with you and takes a little attention every day.
Episode 1 built Coupon, an agent whose job is to stop money leaking out of my accounts. (If the word is new, here is what an agent is.) This episode starts a second team, for health, and its first member exists to hold those loose worries. Its name is Worry Wart. I call it Wart.
What does Worry Wart do all day?
It runs in Grok Bot, SpaceXSI’s tool for building your own agents, the same one from episode 1. The agents can message each other, which matters more with every one I add. To set one up, I clicked a plus sign, typed a name, added a profile picture and pasted one long message. I left the description field blank because the agent writes its own.
Grok Bot is not on the free tier. Its FAQ says it comes with the Premium, Super and Ultra tiers of the X Subscription (Plus and Free do not include it), plus linked SuperGrok or X Premium+ accounts and some older plans. Every plan has a weekly usage limit, larger on the higher tiers. A daily sweep counts against that limit, so I keep the list of jobs short.
Wart has four standing jobs. Each one becomes a routine, Grok Bot’s name for a task that runs by itself on a schedule.
- A daily sweep for recalls and safety alerts. It only counts the severe ones, such as food causing serious illness, consumer electronics catching fire, faults in cars and problems with passenger aircraft.
- A weekly check-in with a diet and nutrition agent on the same team, to learn which foods and brands are in my kitchen.
- A monthly note on real advances in science and medicine, with an instruction to stay quiet in any month where nothing qualifies.
- A twice-monthly scan for products almost everyone uses that people are starting to ask questions about.
What closing a mental tab is worth
Recall notices are easy enough to find. The hard part is remembering to look, on top of all the other half-finished things I feel I should keep up with. Everyone has a few of these.
Once I know Wart is scanning and will message me when something serious turns up, I can stop holding the thought and spend that attention elsewhere. I mostly built it for that relief.
A notice also does not always lead to a fix. In a 2024 report, the Government Accountability Office noted that 69% of the passenger vehicles recalled in 2018 had been repaired within 15 months of a fix being available. That leaves about three in ten not yet fixed at that point.
Why check sources instead of headlines?
The prompt sends Wart to the agencies directly, because news coverage usually follows the notice and tends to make it sound bigger. It names six sources and gives each one a bar to clear before I hear about it. All six are US bodies, so swap in your own country’s equivalents.
| Source | What it covers | Wart only pings me for |
|---|---|---|
| FDA | Most packaged food, medicines, medical devices, cosmetics | Class I recalls |
| USDA FSIS | Meat, poultry and egg products | Class I recalls |
| CDC | Outbreak investigations across states | Open outbreaks with deaths or hospital stays |
| CPSC | Household goods, toys, electronics | Recalls or warnings citing serious injury or death |
| NHTSA | Cars, tires, child seats | “Park outside” and “do not drive” notices |
| FAA | Aircraft | Emergency airworthiness directives |
Four terms in the prompt
- Class I recall
- The most serious of three levels. The FDA defines it as a “reasonable probability” that the product “will cause serious adverse health consequences or death.” USDA uses the same wording for meat and poultry.
- Product safety warning
- What the CPSC usually issues when a seller has not agreed to a recall or a remedy (an example). Its recalls page lists warnings and recalls together.
- Emergency airworthiness directive
- An FAA order issued “when an unsafe condition exists that requires immediate action by an owner/operator.” It is written for airlines and aircraft owners, so a passenger does not need to do anything about one.
- Outbreak
- A group of illnesses that investigators have traced to the same source.
Going to the source has two limits. The FDA says on its own recalls page that not all recalls have press releases or are posted there. And when I rechecked all six pages on November 14 using a simple program in place of a browser, FSIS and NHTSA blocked it with an error. An agent can hit the same wall and say nothing. Ask yours once to list which sources it could read that day.
The prompt also tells Wart what to leave out. It skips recalls for undeclared allergens (an ingredient such as milk or peanuts missing from the label) and other labeling mistakes, unless the product is one I eat. My prompt calls that category “roughly half” of all food recalls, and that was me rounding up. The closest count I have found since is 39%. That comes from the PIRG Education Fund’s Food for Thought 2026 report, which tallied 320 FDA and USDA food recall announcements in 2025 and found undeclared allergens were the most common cause. It is still the biggest single cause, so skipping it removes more alerts than any other line in the prompt. If someone in your house has a food allergy, delete that line.
How long does a recall take to arrive?
There is one deliberate exception in the prompt. For possible salmonella, E. coli or similar contamination, Wart is told not to wait for a formal recall.
When I wrote it I assumed a recall ran a few days behind the first illnesses. It is closer to weeks. The CDC puts the usual reporting lag for a foodborne outbreak at three to four weeks, because someone has to get sick, see a doctor, be tested and have the result matched to other cases. The recall itself adds more time. A government watchdog’s audit of 30 FDA food recalls found that half of them took 29 days or longer from the FDA learning a product might be hazardous to the recall starting. The auditors picked those 30 cases by hand from 2012 to 2015, so the figure shows how slow a recall can be and may not be typical today.
I wrote the exception because of that gap. It also makes this the job most likely to be wrong, because Wart is passing on a suspicion before any agency has confirmed it. When Wart flags possible contamination early, check the agency page yourself before you throw food out, and before you tell anyone a named company has a problem.
How does it get to know you?
A recall sweep is only useful if it knows what is in your cupboard. I could list my brands in the prompt and then go back to update that list forever.
So the prompt tells Wart to ask a teammate. Once a week it checks in with a diet and nutrition agent, pulls what I eat, and refines its own alert list from that. When I recorded this episode that agent did not exist. It does now. Lettuce, from episode 3, keeps my food diary from photos, so Wart got better without anyone editing it. 2nd Med, from episode 4, joined the same team after that. They can ask each other things because they all live in Grok Bot.
The early warning job
Twice a month, Wart looks for products almost everyone uses that people have started to doubt, or that might be doing harm. The example written into the prompt is LED lighting, where the claim is that it disrupts sleep cycles and possibly hormones.
When I typed that, it was hearsay. I had heard it somewhere and not looked it up, and the prompt says as much. I want this job to tell me when a question starts circulating, early enough that I can look into it properly. Most of those questions go nowhere.
A few turn out to matter, usually long after the first doubts. The US banned lead paint for homes in a rule announced in 1977. The EPA finalized its ban on chrysotile asbestos, the last kind still imported, in March 2024. The painkiller Vioxx was withdrawn by its maker in September 2004. For each of those there were many scares that faded, so I treat anything from this job as a question to look into.
So Wart is allowed weaker evidence here than for the alerts. It can read what people are saying on social media, personal stories and early studies, as well as agency publications.
What does the evidence on LED light say?
Since the prompt uses it as the example, I went and looked. The part about hormones holds up in one narrow sense. Melatonin, the hormone that tells your body it is night, drops under bright, blue-rich light in the evening.
- In a 2015 lab study, 12 adults who read on a light-emitting e-reader for about four hours before bed had their evening melatonin cut by about 55% and took nearly 10 minutes longer to fall asleep than when they read a printed book.
- A 2016 American Medical Association report on LED street lighting estimated that a white LED lamp is at least five times more powerful at shifting the body clock (the circadian rhythm) than the older sodium lamps, and recommended warmer, lower-blue lights for streets.
- A 2025 review of blue-blocking glasses found three small trials and concluded the evidence does not show a significant effect on sleep.
None of that shows the LED bulbs in a living room harming anyone over the long run. The first study is about screens held close to the face, and the second is about street lights. So my answer for now is to dim things in the last hour before bed and leave the bulbs where they are. Working that out took me twenty minutes, and I expect to do the same with whatever Wart flags next.
The full prompt
Paste the whole thing into a new bot in one go. Everything in it becomes a standing instruction, so you never have to explain any of it again. Change the six sources if you live outside the US, and read the allergen line before you keep it.
Your name is Worry Wart. I'll call you Wart, and you are part of my health and
wellness team that I'm building. You are the first agent with more to come. Your
role is implied in the name, and it is to do the worrying for me by keeping a
lookout and actively scanning for things that can impact my health, both good and
bad. Here are the things that I want you to automatically and continually do for me:
Be on a daily lookout for recalls and safety alerts. Set your alerts for anything
that is extremely hazardous to my health or deadly. Food that's causing severe
illnesses, consumer electronics that are blowing up, issues with cars, concerns
with passenger planes. Check these sources directly rather than relying on news
coverage, which lags and overstates:
FDA recalls, market withdrawals, and outbreak advisories
USDA FSIS recalls and public health alerts
CDC outbreak investigation pages
CPSC recalls and consumer warnings
NHTSA recall campaigns
FAA airworthiness directives
Only alert me if it clears one of these bars:
FDA or USDA Class I
CPSC recall or warning citing risk of serious injury or death
NHTSA "park outside" or "do not drive" notice
FAA emergency airworthiness directive
Any open outbreak with reported deaths or hospitalizations
Skip undeclared allergens and mislabeling unless it involves something I actually
consume. That category is roughly half of all food recalls and is noise for my
purposes.
Don't wait for a formal recall for anything with possible Salmonella, E. coli or
other serious contamination concerns.
Once a week I want you to check in with the diet nutrition agent on the team (that
will be created in the future), and get details on the brands of food and types of
food that I consume so that you can further refine your alert system for my
lifestyle.
Each time you do your monthly poll where you ask the team questions, I want you to
continue to refine your list of categories for positive developments that I would
care to know about. Same rule as with negative news, I want you to check the
sources directly rather than through science journalism. I want to avoid hype and
exaggeration. Only let me know if these are legitimate breakthroughs. If there's
nothing worthwhile to let me know about for the month, then no need to ping me
about it.
I also want you to do research twice a month to see if there is any widely used
product that has had its benefits doubted, or that may even be harmful. My most
recent example is LED lights, where there appears to be mounting evidence that they
are disrupting our natural cycles and possibly even hormones. For here you can
search broadly through direct sources and social media sentiment, scientific
studies and even anecdotes. This is a jumping off point for me to initiate more
research, so unlike the other bars that need to be cleared for the alerts, think of
this more as an early warning system for me.
What Wart cannot do for you
- It is not a doctor and gives no medical advice. It forwards what agencies publish and what researchers report.
- It can miss things. Not every recall is posted, and some agency pages turn automated readers away.
- An early-warning flag proves nothing by itself. If Wart says LEDs are being questioned, you still have to read the evidence.
- It cannot tell you which risks matter in your life. You still have to decide whether a worry is real or just fashionable.
Free official alerts, if you would rather not build anything
You can get the raw feed without an agent. It arrives unfiltered, so you will get every minor notice along with the serious ones, but it costs nothing and takes a few minutes to set up.
- FDA email updates: free alerts by topic, recalls included.
- CPSC recall emails, plus a recall app that opens on your phone.
- NHTSA’s recalls page: where car, tire and child seat recalls are published.
- Recalls.gov: one portal set up jointly by six federal agencies. It does not include the FAA.
Made for regular people, by a real person
A free daily newsletter about using AI in ordinary life, curated and edited by a human. Plain English, no jargon, no paywall.
Free forever. Unsubscribe anytime.
The series so far
Each episode builds one agent and publishes its full prompt. Read them in order if you can, because later agents talk to earlier ones.
- Episode 1: Coupon, the money agent. Cancels subscriptions, hunts rebates, tracks warranties. Watch episode 1.
- Episode 2: Worry Wart, the health agent. This post. Watch episode 2.
- Episode 3: Lettuce, the nutritionist and chef. Logs meals from photos, tracks supplement doses, reads restaurant menus. Watch episode 3.
- Episode 4: 2nd Med, the medical second opinion. Reads results, flags drift, checks the billing. Watch episode 4.
Common questions about building Wart
Do I need to know how to code?
No. The prompt above is plain English. You click a plus sign, type a name and paste the text. It is about as hard as writing a long message to a colleague.
Do I have to build Coupon first?
They are separate agents on separate teams, so Wart works without Coupon. It gets better once it has teammates to ask.
What if it tells me something alarming and wrong?
Assume it will at some point. Agents sometimes do things nobody asked for. Our newsletter covered a test in which an AI model typed a made-up tip into a police website. Wart can only read and report, so the worst it can do is tell me something wrong, and I pointed the recall job at the agencies to make that less likely. When it pings you, read the notice yourself before you do anything.
Can I use this for something other than health?
Yes. Any team can have its own version. A work team could have an agent watching for what could hurt your business or your industry. Change the sources and how serious something must be before you hear about it, then keep the four jobs.
Why leave the description field blank?
Because the agent fills it in itself once it has the prompt. One less thing to write.
References
- SpaceXSI, Introducing Grok Bot, the Grok Bot FAQ and Skills and routines: the platform, its plans and its schedules (read November 14, 2026).
- FDA, Recalls, market withdrawals and safety alerts and Recalls background and definitions.
- CDC, Current outbreaks and Timeline for identifying and reporting illnesses.
- CPSC, Recalls and product safety warnings.
- FAA, Airworthiness directives and Types of airworthiness directives.
- U.S. PIRG Education Fund, Food for Thought 2026: food recalls and outbreaks in 2025.
- HHS Office of Inspector General, audit A-01-16-01502: FDA’s food recall process, 2017.
- GAO-24-106356: repair rates for recalled vehicles, 2024.
- Chang and others, Evening use of light-emitting eReaders: PNAS, 2015.
- AMA Council on Science and Public Health, Report 2-A-16: LED community lighting, 2016.
- Luna-Rangel and others, blue-blocking glasses and sleep: Frontiers in Neurology, 2025.