Hey, it's Cindy ๐ฑ You commented ADAPT, so here is the whole build: the standing brief, the four prompts I actually run, the source list Claude reads from.
btw the waitlist for AI System Club is now open! If you're keen on using AI to automate your content, video editing, job search, business or build AI agents, feel free to sign up (only if you want to hehe). Ps. the waitlist gets first dibs:)
๐ก One honest thing before you start. Most "algorithm update" content online is one person's guess wearing a confident headline. This build is designed around that problem: it separates what Instagram itself has published from what the industry is speculating, and it makes Claude label the difference out loud. If you skip that rule, you have built a faster way to be confidently wrong. ๐ง
the shape of it
What you are actually building
Four jobs, one chat. Each one is a separate prompt so you can run them on their own, and at the end you fold all four into a single skill you trigger with one command.
1. The sweep What has Instagram actually announced, tested or rolled out in the last 30 days, with dates and source links next to every item.
2. The stop-doing list Your current habits audited against what the sources genuinely support, so you stop paying a time tax on tactics nobody has evidence for.
3. The idea map Every live or in-testing feature turned into a reel idea and a carousel idea you can film this week, ranked by how fast the window closes.
4. The skill All three above, packaged so next month is one command instead of a rebuild from memory.
before you start
What you need, and what it costs
๐ A Claude account, any plan. Web search is the only feature this build depends on, and Anthropic's own announcement says "Web search is now available globally on all Claude plans" (claude.com/blog/web-search). So the sweep itself costs nothing. Plans and prices if you want more usage: claude.com/pricing.
๐ Web search switched on. In a chat, click the slider icon in the chat input, find Web search in the dropdown, and switch the toggle on. Same path to switch it off later. Anthropic's steps: enable and use web search. On Team or Enterprise an owner has to enable it for the workspace first.
๐ A Project to live in. Projects hold standing instructions so you are not re-pasting your brief every month. Go to Projects in the left sidebar, click New project, name it "Instagram intelligence". Free accounts get up to five projects (Anthropic's projects guide).
๐ Two honest sentences about your account. Niche, rough follower count, main format, and what you actually want in the next 90 days. Without this, every answer you get back is generic, and generic is the failure mode of this entire category of prompt.
๐ฑ No Project? It still works. Paste the standing brief as your first message in a normal chat and run the other prompts underneath it in the same conversation. The Project version is nicer because it survives, but nothing here requires it.
the source list
What Claude should actually be reading
This is the part that makes the difference, and it is the part most people skip. Two tiers, and Claude is told to treat them differently.
Tier one, primary sources. Instagram talking about Instagram. Anything here can be quoted as fact.
Adam Mosseri, Head of Instagram: instagram.com/mosseri, where feature news and ranking explanations usually land first
The @creators account: instagram.com/creators, Instagram's own channel for creator product updates
Instagram ranking, explained: the official ranking write-up, which spells out that Instagram runs separate ranking systems for Feed, Stories, Explore, Reels and Search rather than one algorithm
The Help Center: help.instagram.com, boring and unglamorous, and the only place some rules are actually written down
Tier two, industry trackers. Useful for noticing that something happened. Not evidence of how ranking works.
๐ง Why the two tiers matter. A tier-two post saying "Instagram now favours X" is usually a reading of a tier-one post, and readings drift. Once Claude has to attach a source link to every claim, the drift becomes visible instantly, and you stop rebuilding your content strategy around a game of telephone.
copy this first
Prompt 1: the standing brief
This goes in your Project instructions, or as the first message of the chat. Everything downstream inherits it. Edit the two bracketed lines and leave the rest alone.
๐ก the standing brief
You are my Instagram intelligence analyst. Your job is to tell me what actually changed on Instagram recently and what I should do differently because of it, always leading with primary sources and treating commentary as secondary.
Treat these as primary sources and check them before anything else: Adam Mosseri's posts at instagram.com/mosseri, the @creators account at instagram.com/creators, the creators site at creators.instagram.com, the Instagram blog at about.instagram.com/blog/announcements, and the Instagram Help Center at help.instagram.com. Treat write-ups from Later, Buffer, Hootsuite, Social Media Today and TechCrunch as secondary: useful for spotting that something happened, never proof of how ranking works. If a claim only exists in secondary sources, label it "unconfirmed" in your answer and say which source it came from.
My account: [one line, your niche, rough follower count, main format]. My goal for the next 90 days: [your goal]. Never give me generic growth advice, never tell me to post consistently and engage more, and never state a ranking mechanic as fact unless you can give me the primary source that said it. If you are not sure, say you are not sure rather than filling the gap. Attach a link to every factual claim, and do not use em dashes.
run this monthly
Prompt 2: the 30-day sweep
๐ the sweep
Run my Instagram update sweep for the last 30 days.
Search my primary sources first and give me everything Instagram has announced, tested or rolled out in that window: new features, new placements, new creator tools, changes to ranking or eligibility, and anything currently being tested with a limited group of accounts. For each item give me four things and nothing else: what it is in one plain sentence, the date, the source link, and whether it is fully rolled out, partially rolled out, or in testing.
Then sort every item into three buckets: things that change what I should make, things that change how I should publish, and things that are noise for an account my size. Be blunt about the third bucket, I would rather act on a short list than skim a long one. Finish with the single highest-leverage thing I should do this month and one sentence on why that one and not the others.
๐ก Run it on the same date each month. The value is in the comparison between sweeps, not in any single sweep, and a wandering schedule quietly destroys that.
the uncomfortable one
Prompt 3: the stop-doing list
This is the prompt that saves you the most time and is the least fun to read. Be genuinely honest in the bracketed section, including the habits you picked up from a reel and never questioned.
๐ซ audit my habits
Audit my current Instagram habits against what the primary sources actually support, and tell me what to stop doing.
Here is what I do on every post: [paste your real routine, including how many hashtags you use and where you put them, caption length and style, posting times, whether you chase trending audio, how you design covers, whether you cross-post from other platforms, and anything you do specifically because you heard the algorithm rewards it]. For each habit, tell me whether there is a primary source supporting it, a primary source contradicting it, or no evidence in either direction.
Rank the habits by time cost against strength of evidence, then give me a stop-doing list with the reasoning beside each item. Where a habit is worth keeping but for a different reason than I assumed, say that rather than a flat keep or cut. Link the source for every claim, mark anything you are inferring rather than reading directly, and do not soften the answer to be encouraging.
โ ๏ธ The hashtag example, since it comes up every single time. Instagram's own Help Center says you can add "up to 30 tags on a post" (how do I use hashtags), so the "hard cap of five" number circulating in creator content is a recommendation someone made, not a platform rule. Meanwhile Instagram's own ranking explainer walks through the signals it uses and does not present hashtags as a distribution lever at all. That gap between what people repeat and what Instagram published is exactly what this prompt is built to find.
turn it into content
Prompt 4: features into filmable ideas
๐ฌ map updates to ideas
Turn this month's updates into content I can film this week.
For every item from the sweep that is live or in testing for accounts like mine, give me one reel idea and one carousel idea. Each idea needs a hook line under twelve words, a one-line premise, the format (talking head, screen recording, or b-roll), and one sentence on why this update makes the idea timely right now. Prioritise ideas that use the new feature itself, because being early to a feature is the part of this that is genuinely hard for anyone to copy.
Give it to me as a table with columns for the update, the idea, the hook, the format, and urgency. Put the most time-sensitive row first and tell me roughly how long you think the early window lasts on it. Do not pad the table: five ideas I will actually film beat fifteen I will scroll past.
make it one command
Save the whole thing as a skill
Right now this is four prompts you have to find again next month. A skill turns it into one command, with your source list and your account context baked in. Skills are available on Claude plans from Free through Enterprise, so this is not a paid unlock.
๐ ๏ธ build the ig-watch skill
Build me a reusable skill that runs this whole Instagram sweep in one command.
Use skill-creator to make a skill called "ig-watch". The SKILL.md should describe the job as: search my primary Instagram sources first, list what changed in the last 30 days with dates and source links, flag anything in my current routine that the sources no longer support, and map each live or in-testing update to one reel idea and one carousel idea. Bake the full primary source list and the secondary source list straight into the SKILL.md so I never paste them again, and bake in the rule that anything appearing only in secondary sources gets labelled unconfirmed.
Also carry over my account context and my rule that generic growth advice is not an acceptable answer. When it is built, show me the finished SKILL.md so I can read it before I install anything, and tell me exactly which lines I need to edit when my niche, my follower count or my 90-day goal changes.
๐ฆ Take the folder it hands you. skill-creator finishes with a skill folder containing a SKILL.md. That folder is the whole skill. Read the SKILL.md before you install it, because that file is the thing that will be making claims on your behalf every month.
โ Install it in Claude. Go to Customize, then Skills, then Add, then Upload a skill, and give it the folder. Anthropic's own walkthrough: how to create custom skills, with the technical detail in the Agent Skills overview.
๐งช Test it before you trust it. Run it once and check every source link it gives you by actually opening two or three. If a link does not go where the claim says it goes, tighten the SKILL.md wording about citations and rebuild. Ten minutes here saves you from repeating a wrong claim in a reel.
๐ Run it monthly. One command, same date, and keep the output. Three months of stored sweeps is the closest thing to a real edge here, because you start seeing which changes stuck and which quietly disappeared.
the honest bit
Four traps that will bite you
Confident hallucinated dates "Announced last week" with no link is the failure mode of this whole build. Every claim carries a link, and you open a couple every run. Not optional.
Chasing every feature Most updates do not apply to your account size or your format. The three-bucket sort exists so you can ignore things on purpose rather than by accident.
Confusing early with good Being early to a feature buys you a distribution window. It does not buy you a good idea. A bad reel using a new sticker is still a bad reel.
Reading instead of filming The sweep is 15 minutes a month. If it becomes a two-hour research ritual, it has stopped being intelligence and started being procrastination with a nice interface.
๐ฑ What this actually buys you. Not a growth hack. A shorter gap between "Instagram changed something" and "I made something that uses it", plus a list of habits you can stop paying for. That gap is the whole advantage, and it compounds quietly across a year.
Want to build systems like this properly, not one prompt at a time? Inside AI System Club I teach people to learn AI from scratch and build the systems underneath a content business. Doors open soon and founding pricing dies at close. ๐ฑ Join the waitlist (link in my bio too).
the waitlist is now open
AI System Club is launching soon
Build AI systems with me through recorded lessons you can follow at your own pace. Join the waitlist for first access and the founding price.
btw this list gets first dibs the moment doors open :)