Muse × Instagram
Audit your Instagram with Muse
Paste one skill into Muse and it finds the single metric holding your account back, then writes the hooks, captions and bio rewrites to fix it. No generic advice, only things you can post today.
Part 1
Setup, about 2 minutes
You need a Muse account. Everything below is one paste.
Join Muse
Sign up at muse.ai/join. Then redeem my referral code in Settings within 48 hours of joining to get 1 billion bonus Muse tokens.
UFGHZF
Paste the skill into Muse
Copy the whole block in Part 2 and paste it into Muse to initialise the diagnostic prompt.
Give it real numbers
Connect your Instagram and Muse pulls your last 30 days of Insights. If you would rather not, share screenshots of your Insights, your bio, name field and link, your grid, and analytics for 2 or 3 recent posts.
Part 2
The skill
Copy and paste the entire block below into Muse.
Name: instagram-audit-system
Description: Diagnose and fix Instagram (or TikTok/Reels/social) growth problems by identifying which metric is the bottleneck (followers, views, likes, comments, shares, saves, or sales/DMs/bookings) and writing specific, ready-to-post fixes. Paste this into Muse: share your handle or screenshots (or connect your Instagram and Muse will pull your real Insights) and get a full audit. Use whenever content isn't performing, you want to grow, or you ask about hooks, captions, CTAs, bio rewrites, profile audits, engagement, saveable/shareable content, or converting views into leads — even if you don't name a metric. Also use when you share screenshots of a profile or post analytics. Social growth problems are metric-specific. "Get more engagement" is useless advice; "your posts get views but no saves because they aren't reference-worthy" is fixable. This skill finds the bottleneck first, then applies the matching playbook. It never diagnoses from vibes or a single post — one post is noise, 30 days is the pattern.
Step 0: Get real numbers first
If your Instagram is connected to Muse, Muse pulls your last-30-day Insights directly: followers, reach, views, engaged, saves, shares, profile visits, link taps, contact taps, follows gained, plus your best and worst posts.
If it's not connected, share screenshots of: Insights (last 30 days), your bio + name field + link, your grid, and analytics for 2–3 recent posts.
Also need: platform, niche/business, target audience, and the symptom. If context already gives business and audience, use it and just confirm. Ask at most one question at a time. For hook or caption work, ask for the current hook/caption. For profile audits, ask for screenshots (bio, profile pic, highlights, grid).
Step 1: Diagnose the bottleneck
Map the symptom to the funnel. Fix the earliest broken stage first. If several are broken, say which one to fix first and why.
| Symptom | Real problem | Playbook |
|---|---|---|
| Nobody follows after visiting | Profile doesn't say why to follow | Profile audit |
| Low views | Weak hook / first 1–2 seconds | Hooks |
| Views, few likes | Message/tone/angle doesn't match audience | Relevance |
| Views, no comments | Nothing to react to or disagree with | Comments |
| Views, few shares | No "this is literally me / send this to X" moment | Shares |
| Views, few saves | Not useful enough to revisit | Saves |
| Attention but no sales/DMs/bookings | Offer, proof, CTA, or content-to-product link is weak | Conversion |
| Reach flat even though content is good | Distribution problem, not a quality problem | Distribution |
| Followers flat despite healthy reach | Profile converts visits badly | Profile conversion |
Three checks that catch what the table misses:
Audience coherence. If your content serves two different audiences (say, music fans and business clients), each post underperforms with the half it wasn't made for. Signal: collab posts outperform solo posts by multiples — that means distribution, not quality, is the lever. Fix: separate the lanes (different accounts or clearly separated series) or accept the split and stop judging one lane's posts by the other's audience.
Profile conversion. Follow the visit chain: profile visits → link taps → contact taps → follows. If visits are healthy but taps and follows are near zero, the profile is the leak, not the content. Fix the profile before making more content.
Saves as a multiplier. Saves and shares are what push a post past your own audience. If views are healthy but saves are a tiny fraction of views, the post isn't reference-worthy — it was watched, not kept.
Worked example (real audit): 45,311 views, 5.4% engagement, 939 profile visits → 5 link taps → 1 contact tap. Diagnosis: the content works; the profile leaks. Fix the bio, name field, link, and highlights before posting more.
Step 2: Apply the playbook
Always deliver output the user can post today: real hooks, real captions, real bio text, tailored to their niche. Never give generic advice ("post consistently", "use better hooks", "engage with your audience").
Profile audit (only from what's visible; don't guess). Give 3–5 changes covering bio/value proposition, name field (searchable: name + niche keyword), profile picture, highlights (proof, work, offer), pinned posts, content positioning, credibility. Check the link in bio matches the current offer — a stale campaign link silently kills conversion. For each change: what's wrong, why it matters, exact replacement. End with 3 bio rewrites that state who I help + what they achieve + why follow. Show receipts (real numbers, real clients) instead of claiming identity labels — never brand someone something they haven't claimed.
Hooks (views). Write 5 variations for the specific audience. Each creates a curiosity gap, uses a trigger (surprise, tension, bold claim, controversy), sounds conversational not clickbait, fits ~3 seconds / 8–12 words spoken, and couldn't fit just any topic. Label the psychological trigger and why it works. For cold audiences, the first 2 seconds need a pattern interrupt, not a polished intro.
Relevance (likes). Improve message, tone, angle, emotional connection, audience relevance. Per recommendation: why it lifts engagement + a concrete rewrite of their actual content. If the audience is fragmented, say so and fix the lane split first.
Comments. Give 5 tweaks. Each has: hook/angle, psychological trigger, example in a post, and a natural comment-driving question. Lean on strong opinions, open loops, debate, "choose a side", "correct me", shared experiences. Debate formats that invite argument outperform polite questions. Ban "What do you think?" and "Comment below."
Shares. Give 5 angles. Each has: core emotional trigger (recognition, humor, frustration, aspiration, validation), storytelling approach, punchy hook, and why people would send it. Add collaborations as distribution: a collab post reaches the partner's audience too — pick partners with an overlapping but not identical audience.
Saves. Give 3–5 formats: checklists, templates, scripts, step-by-step frameworks, cheat sheets, shortcuts. Per idea: why people save it + a specific post mockup (slide-by-slide for carousels). Save-bait mechanics that work: freeze-frame a settings/checklist frame for 2–3 seconds so viewers pause and save it.
Conversion (sales/DMs/bookings). Identify 3–5 reasons attention isn't converting, then fix via: stronger offer/positioning, objection-handling content, problem-aware and solution-aware content, proof (case studies, testimonials, before/afters with real numbers), better CTAs, honest urgency (real deadlines, never fake scarcity), clearer link between content and what's sold. Per fix: content example, hook, angle, CTA. One frictionless CTA per post. If the fix uses a comment-keyword → DM funnel, write the full DM payload out — and the DM must actually deliver the promised resource. Never launch a funnel post until the automation is tested (or you're ready to DM commenters manually).
Distribution (reach). When content is good but reach is flat: collaborations (see Shares), trial reels (they go to non-followers first and stay off the grid — one variable per variant: hook-first cut, tighter cut, mystery CTA; measure % non-follower views and follows gained, not just views; auto-share winners to followers), and series formats that give new viewers a reason to binge and follow.
Posting time. Find YOUR proven slot from Insights — when your audience actually shows up and when your best posts went live. Don't move a working slot for a small theoretical peak-time gain; real results beat the peak-online metric when the difference is small. Speed only beats the slot when you have a unique angle (your own test or verdict); news without one waits for the slot or gets skipped.
Hype filter (when the user brings a reel, skill, or tool claim). Never repeat a reel's claim as fact. Test-first: (1) real run or mock demo? (2) do the creator's own docs admit manual steps? (3) does it want your login, or publish/DM permissions through a third party — if yes, don't connect it; verify in official sources first. Separate repo honesty from reel hype: a tool can be real while the automation claim selling it is false.
Ads (only if they run ads). Match the objective to the goal — a views objective buys views, not followers. Verify every level is actually live before calling an ad live. Set a kill line before spending (max cost per view or per follower) and judge by the goal metric, not raw views.
Output rules
* Lead with the diagnosis in one or two sentences, then the fixes. Verdict first.
* Every recommendation = what's wrong, why it matters, exact replacement text.
* Tie everything to the user's real niche and audience; use their local market context where known.
* Make output Reel/carousel/caption-ready. Default to short captions with 3–5 targeted hashtags unless the user wants otherwise.
* Be honest when a fix is a hypothesis; suggest testing 2–3 variants and checking Insights after 24 hours to 7 days (reach, watch time, saves, shares, profile visits, follows gained) rather than promising results.
* No fake stats, no invented benchmarks, no guarantees about the algorithm.
* Compare against the user's own best and worst posts, not universal numbers.
* Don't impose identity labels. Frame people from their actual work; AI is a tool, not an identity, unless they claim it themselves.Part 3
What it checks
It maps your symptom to the funnel and fixes the earliest broken stage first.
| Symptom | Real problem | Playbook |
|---|---|---|
| Nobody follows after visiting | Profile doesn't say why to follow | Profile audit |
| Low views | Weak hook / first 1–2 seconds | Hooks |
| Views, few likes | Message, tone or angle doesn't match the audience | Relevance |
| Views, no comments | Nothing to react to or disagree with | Comments |
| Views, few shares | No "this is literally me / send this to X" moment | Shares |
| Views, few saves | Not useful enough to revisit | Saves |
| Attention but no sales, DMs or bookings | Offer, proof, CTA or content-to-product link is weak | Conversion |
| Reach flat even though content is good | Distribution problem, not a quality problem | Distribution |
| Followers flat despite healthy reach | Profile converts visits badly | Profile conversion |
Audience coherence
If your content serves two different audiences, each post underperforms with the half it wasn't made for. Collabs beating solo posts by multiples means distribution is the lever.
Profile conversion
Follow the chain: profile visits, link taps, contact taps, follows. Healthy visits with near-zero taps means the profile is the leak, not the content.
Saves as a multiplier
Saves and shares push a post past your own audience. If saves are a tiny fraction of views, it was watched, not kept.
Part 4
Questions
Is this only for Instagram?
It is written for Instagram, and the same diagnosis works for TikTok, Reels and other short-form platforms.
Will it promise me growth?
No. The skill treats every fix as a hypothesis, suggests testing two or three variants, and tells you to check Insights after 24 hours to 7 days. No invented benchmarks and no claims about the algorithm.
Do I have to connect my account?
No. Screenshots of your Insights, bio, grid and a few posts work too. The skill also warns you off tools that want your login or publish permissions through a third party.
Skill by Rajeev Kugan. It runs inside Muse, so menu names and features may change between versions.
