agentleFS
Evan-Kim2028Evan-Kim2028/agent-skills/skills/instagram/instagram-virality-insights/SKILL.md

instagram-virality-insights

Skill1 starsChanged 3 months agoMITView on GitHub

Distilled virality and retention mechanics for short-form data reels (open cognitive loops, stakes/urgency framing, share psychology, duration bands, frame-0 mechanics) with measured 2026-08 evidence. Use when designing new reel concepts/series, judging why a reel or account travels or stalls, or pressure-testing an episode idea before production. Optional live-recon appendix for gathering fresh evidence. Not for production ship-gates (high-retention-insight-reels-instagram owns those) or offline .mp4 review.

The file

We split this skill into the rules it gives an agent

1 rule. About 0 are common, 1 is distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈1

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    OPTIONAL: Use only when fresh evidence is needed, not routinely — recon findings decay and one session's sample is directional, not proof.

    line 76

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Instagram Virality & Retention Insight Quality

Core law: the open cognitive loop

Reels travel when frame 0 opens a question the brain must chase; they stall when frame 0 is resolved. Confrontation (belief attack) is one loop type, not the only one. A finished chart = no loop = the measured 0.003–0.09 v/f band (CardLadder/GemRate).

Loop familyMechanic
Belief attack"You're wrong about X"
Unresolved wagerOutcome revealed at the end
Closing windowUrgency / deadline
Rank churnRace, winner hidden
Scale violationTrue-size "whoa"
Crisis / news"What happens to us now?"

Evidence table (2026-08, measured)

WhoSizeResultMechanic
thecardscience56.6K135K views in 3 daysPrice dashboard + "discounts don't last forever" urgency
thecardscience (same acct)56.6K2.2M pinnedTarget-cancels-Pokemon crisis news
legalrips693 followers1.2–5K every reelReal-money coinflip resolving on camera
card_dons—515 comments / 184 repostsDirective covers ("BUY THIS AVOID THAT")
chartosaur239K47–149K consistent, zero flopsTemplate-consistent tension, 14–20s
onlyslabs.io15 followers5–162 viewsNegative control — slab data, no stakes
Silph July batch—12/21 confrontation-worded hooks died at ~150–215 viewsWording ≠ mechanism; frame-0 visuals were calm
Silph Gengar (Jul 23)32 fol1,687 views, 11 shares, 99.3% non-followersUnresolved set fight on cover (DIDN'T CRASH); cascade proofs
Silph gold-delta (Aug 26)32 fol7–145 views with send-asks + WRONG stampResolved cover — punchline on the thumbnail

Retention mechanics (ranked by evidence strength)

  1. Duration — 12–25s for data reels; 31s+ = measured losing band. Length never fixes a resolved open.
  2. Familiar-entity-first — a recognizable card/entity IS a hook before any claim text.
  3. Payoff cadence — ~3–5s; text cards that ARE the story ≠ dead air. True dead air = logo stings / empty wipes.
  4. Loop seam — last ~2s should visually rhyme with first ~2s (replays count as watch time).
  5. Data as evidence inside a story with stakes, never as the subject.

Share/save psychology

  • Sends-per-reach is the top expansion signal (Mosseri-confirmed). This is a caption-layer pattern (Youshaei hook-body-send-ask; Mosseri sends-per-reach), not a VO pattern — the send-ask is mandatory but lives in the caption (and optionally a pinned comment), never spoken or on-screen in the final beat.
  • Recipient-named send-asks ("send this to the friend who still…") beat quota asks, and beat polls (polls → comments = weakest lever).
  • Saves come from a portable rule or reference numbers held static ~2s.
  • Directive framing ("buy this, avoid that", "hold three years") gets reposted; observations don't.
  • Identity/debate bait (era vs era, which side are you on) drives tagging.

Escaping the test pool (small accounts)

  • Scoring is per-video, not per-account: the first pool measures skip + sends + completion regardless of follower count.
  • Engineer one save reason AND one share reason into every reel.
  • Trial Reels need 1,000+ followers — check before recommending.
  • Collab leverage: data brands travel by powering a personality's reel (double.holo pattern) — SilphCo-as-source is a valid distribution play.

Anti-patterns

Resolved frame 0 (calm green stat) · analytics as subject · poll-only CTA ·

30s data recaps · polish iterations inside a dead template · judging hooks by wording without checking the muted frame-0 visual.

Optional appendix: live recon protocol

OPTIONAL. Use only when fresh evidence is needed, not routinely — recon findings decay and one session's sample is directional, not proof.

Goal

Find organic winner Reels on live Instagram outside the follow graph, vet whether a view count is real (v/f) or a fluke, and extract the frame-0 mechanic + CTA driving it.

Protocol

StepAction
1. Keyword searchOpen https://www.instagram.com/explore/search/keyword/?q=<query> — top results are IG's own ranking for the topic, not the follow graph
2. Open winner reelsClick into top results, note the account
3. Vet the accountOpen /<handle>/reels/, read "View Count Icon" values vs follower count. Consistent v/f >1 at small scale beats one outlier
4. Extract frame-0Screenshot frame-0 of winners (that IS the muted scroller's view); read caption/comments for CTA + debate patterns
5. Negative controlsAlso record stuck accounts (data-without-stakes accounts sitting at 5–200 views) — these disprove "just post data"

Practical notes

NoteWhy it matters
Web Insights hide skip/watch-timeOnly the mobile app exposes those; web/desktop shows views/interactions/shares/saves
Grid views are lifetime, not recentCheck post dates before treating a view count as fresh signal
Snapshots may need a 3–4s waitLet the reel render before screenshotting frame-0
Feed (/reels/) is mostly general-viralSearch beats feed for niche recon
Don't interactNo likes/follows — keep the account's graph clean for repeat recon

Findings template

accountfollowersview bandoutlierframe-0 mechanicCTA type

Relationship to other skills

SkillRole
high-retention-insight-reels-instagramConsumes these mechanics for create/review/rank; owns offline .mp4 storyboard review and ship-gates
instagram (hub)Routes here for concept design and why-it-travels judgments
marketing-going-viralMulti-platform Gold/Silver/Bronze research, not Instagram-specific

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.