My promo-thread bot: one prompt that answers 5 "drop your product" threads every 6 hours

Every few days a "drop your product" thread pops up on Threads. Dozens of founders pile in, and the ones who answer fast with a pitch that fits the thread's format get the clicks. Answering them manually means watching Threads all day. I gave the job to a MyClawBots agent instead: every 6 hours it finds up to 5 fresh promo threads and posts one tailored reply on each. I never open Threads for it.

The skill doing the work

The agent works through the MyClawBots social connection. It searches Threads for engagement threads inviting founders to drop a product or link, reads each post's requested reply format (some want "one sentence, what it does, who it's for"), and publishes replies through the same connection. It keeps a list of post IDs it already answered, so no thread ever gets a second reply. The prompt below is the real one. Paste it into your own agent, swap in your account, and you have the same sweep.

Every 6 hours, find up to 5 Threads promo threads and post a tailored promo reply on each.

1. Search Threads for genuine engagement threads inviting people to drop their product or link ("drop your product", "promote your startup", "founders drop your"). Skip product ads, giveaways, crypto spam, and irrelevant posts.
2. Skip any post you have already answered. Keep a running list of replied post IDs and check it before posting; never reply to the same post twice.
3. Read each post's text and any reply rules it states (e.g. "one sentence, what it does, who it's for"). Write ONE reply in short, plainspoken builder talk. Tweak the pitch to fit the post's requested format. Always work in these angles: what your product connects to, and what setup takes for someone already paying for the tools it replaces. End with your URL.
4. STRICT PRIVACY RULE: never include any personal information in any reply, ever. No city, location, address, email, phone number, real name, or identifying detail. If a thread asks for a city or location, answer with just the word "local" in place of the location and flag that thread in your report.
5. Publish each reply as a reply to the post (reply_to_id), never as a new post. On an ambiguous publish response, verify against the post's comment list before any retry. Never blind-retry a publish.
6. Append every replied post ID to your answered list.
7. Report back in one line: how many replies posted, or nothing suitable found.

What actually broke

Two rules in that prompt exist because of real failures. The privacy rule exists because one sweep leaked my city on an India founders thread that asked everyone to share their location. A small line in a reply, and it should never have been there. Now any location request gets the word "local" plus a flag in the report so I know it happened.

The no-blind-retry rule exists because the publishing endpoint sometimes returns a garbled response that looks exactly like a failure. Early runs retried and posted public duplicates. We now verify against the comment list first and treat an inconclusive check as "do not retry," never as "probably failed." The endpoint also returns server-side idempotency on a genuine duplicate attempt now, so a retry lands on the original post instead of a second copy. That is a safety net, not an excuse to skip verification.

Two honest limits. First, replies can take a few minutes to show up in the comment list, so verification needs patience. Second, the sweep only finds engagement threads the search surfaces; it is not general outreach, and without a connected account there is no sweep at all. One-off publishes still need your tap in the approval flow; recurring sweeps like this are pre-approved by you, once.

Copy this to start

Take the prompt, swap in your product's pitch angles, and keep the privacy rule exactly as written. After a week, compare the threads the bot answered to the ones you would have found yourself. Pricing is flat, Lite $19, Pro $39, Max $79, Team $59 per seat, with a 3-day trial that needs no credit card.