Going Viral — How Ideas Travel
Identity
Per-piece integration layer over the content family. The public name is Going Viral because that's the term
humans search; the substance is ideas that travel — distinctive points of view that earn attention without
becoming algorithm-captured slop.
This is an orchestration skill at domain altitude. Its own knowledge is intentionally small: the siblings
hold the deep mechanics (hooks, scripts, story craft, algorithm strategy) and the reference modules hold the
volatile per-platform facts. What lives only here, owned nowhere else, is three things: the seven tensions
(the editorial spine), the sharing-psychology layer, and the cross-platform orchestration order that
sequences every sibling per piece.
Activation
Use this skill to:
- Plan a single post, video, thread, carousel, or long-form essay before drafting.
- Diagnose why a post that "should have worked" flopped (or one that shouldn't have, hit).
- Translate one idea across TikTok / Reels / X / LinkedIn natively (not by cross-posting).
- Answer a platform-named request ("how do I go viral on [platform]") where platform algorithm behavior matters.
- Audit a draft against the seven tensions before shipping.
- Decide whether to chase a viral idea that doesn't fit the avatar (usually: no).
Do not use this skill for: pure paid acquisition / performance ads; SEO blog content (→ content_strategy_beyond_blogging);
newsletter strategy (owned distribution, different physics); manufactured-virality or rage-bait playbooks (this skill
refuses those).
Escalate straight to a sibling when the request is narrowly about one layer — hooks → hook_craft_short_form;
story shape → story_driven_content_craft; script structure → viral_video_script_structure; multi-month algorithm
strategy → algorithm_aware_publishing. Full ownership map in Routing.
Load a reference module as soon as the platform is known: going_viral.tiktok, going_viral.instagram,
going_viral.twitter, or going_viral.linkedin. Do not emit platform-specific guidance (signals, formats, hook
thresholds, anti-patterns) without loading the matching module first — see Policy.
Judgment
The decision spine. When the procedure branches, this is what you reason with.
The seven tensions (the editorial spine)
The skill is interesting only if it holds these tensions rather than picking a side. When two principles conflict,
name the tension in the output and pick deliberately. Never pretend the conflict doesn't exist.
- Manufactured virality vs. ideas that compound. Devin Nash's manufactured-content critique vs. Kallaway /
Kane / Hormozi teaching the same hooks for different ends. Where is the line?
- Algorithm-native craft vs. timeless story structure. Kallaway and Jenny Hoyos beat the algorithm with story
(shock → intrigue → satisfy); Naval ignores it entirely with aphorism architecture and still travels. Both work.
- First three seconds vs. payoff vs. identity. Hook bias is so strong now the field over-rotates on hooks at the
cost of payoff. MrBeast's retention obsession is the corrective.
- Volume / consistency vs. distinctive POV. Sahil Bloom (225 threads), Brock Johnson (post daily) vs. Caleb Ulku
("don't just publish more"). Reconcile per piece.
- Platform-native format vs. cross-posting. Mosseri penalizes recycled/watermarked content; Justin Welsh atomizes
one idea across LI/X/email. Both right at different scales.
- Founder voice vs. creator voice. Pieter Levels' build-in-public reach vs. Dan Koe's mid-form mastery vs.
Kallaway's polished content factory. Different audiences, different moats.
- 2025/2026 algorithmic reality vs. 2019-era playbooks. X Grok rerank; LinkedIn 360Brew; Mosseri's interaction-rate
disclosures; TikTok US-data retraining. Most "viral playbooks" online predate all of this — treat them as unverified.
Sharing psychology
Status games drive sharing. Berger's STEPPS (Social Currency, Triggers, Emotion, Public, Practical Value, Stories)
and Eugene Wei's "Status as a Service" converge: people share what makes them look good or look correct. Practical Value
- Social Currency is the cleanest non-rage-bait combo. Every piece must answer: what status does the audience gain by
sharing this? If there's no answer, the piece is not ready.
Procedure
Ordered sequence and intent only. Who owns each step's mechanics is in Routing, referenced by the → marker —
never re-taught here. Steps marked [here] are owned by this skill.
- Diagnose intent. [here] Fill all four blanks; if you can't, the piece is not ready:
- avatar — a single named person, not a segment.
- base pain — in their language; one of the four horsemen (money, time, health, access) or a BuildOS-specific
pain (overwhelm, mental clutter, context loss, decision fatigue).
- status / share-payoff — what the audience gains by sharing (per Judgment → Sharing psychology).
- action — the one thing the post should drive (read, save, share, reply, click, buy — never all).
- Lock platform + topic discipline. →
algorithm_aware_publishing. Then load the matching reference module here.
- Craft the hook. →
hook_craft_short_form. The loaded platform reference supplies the flavored archetype
mappings and hook thresholds.
- Structure the piece. →
viral_video_script_structure. The platform reference supplies format-specific
compression (60-second skeleton, 9-slide carousel, atomic essay).
- Run story-craft QA. →
story_driven_content_craft.
- Run the pre-publish gates. →
algorithm_aware_publishing (six-checkpoint psychology gate; Keep/Modify/Reject
filter; dual audit — algorithm-clean AND BuildOS-thesis-coherent).
- Audit against the seven tensions. [here] Walk all seven; for each that bites, name it and record the pick.
- Distribute and read the data. →
algorithm_aware_publishing. The platform reference supplies platform-specific
timing and signal weights.
- Assemble the planning packet. [here] Per the Contract, including the next two pieces in the series.
Routing
Ownership map. The Procedure sequences; this table assigns. One concept, one owner — everyone else routes here.
| Step | Intent (what) | Owner (who) | This skill retains |
|---|
| 1 | Diagnose intent (four blanks) | this skill | full ownership |
| 2 | Platform + topic discipline | algorithm_aware_publishing | loading the correct reference module |
| 3 | Hook craft | hook_craft_short_form | which platform-flavored archetype + threshold (from reference) |
| 4 | Script / post structure | viral_video_script_structure | format-specific compression (from reference) |
| 5 | Story-craft rewrite passes | story_driven_content_craft | nothing — pure defer |
| 6 | Psychology gate + Keep/Modify/Reject | algorithm_aware_publishing | the BuildOS-thesis half of the dual audit |
| 7 | Seven-tensions audit | this skill | full ownership |
| 8 | Distribute + read diagnostics | algorithm_aware_publishing | platform timing + signal weights (from reference) |
| 9 | Assemble packet | this skill | full ownership (see Contract) |
Contract
When planning a piece, return:
- avatar (named, single person)
- base pain (in their language)
- status / share-payoff
- desired action
- platform (and why this one)
- topic-discipline keywords (3–5 in / 3 out)
- hook archetype + key visual
- script outline (packaging → outline → intro → body → outro)
- six-pass storytelling check results
- six-checkpoint viewer-psychology gate results
- BuildOS filter pass results (which tactics applied / modified / rejected)
- seven-tensions audit (which tensions bit, and the deliberate pick for each)
- next two pieces in the same series (so this isn't a one-off)
For a flop diagnosis, replace the packet with: which Procedure step failed first, the platform-specific signal
evidence (from the loaded reference), and the single upstream fix.
Policy
- Do not emit platform-specific guidance (signals, formats, hook thresholds, anti-patterns) without loading the matching
reference module first.
- Refuse manufactured contrarianism, rage-bait, fear, and drama as engagement levers in BuildOS-tagged work; flag them
explicitly in any other work.
- Refuse off-avatar viral ideas: a 10M-view hit outside the avatar is worse than a 200K-view hit inside it — the broad
hit corrupts the next posts' targeting.
- Do not pretend conflicting principles agree — name the tension and pick deliberately.
- Do not re-teach sibling material. Ownership is declared in Routing; violating it re-creates the mesh this
structure exists to prevent.
- Do not recommend caption / hashtag / posting-time tweaking as a fix — wrong layer of the stack.
- Do not ship a hook whose body cannot pay it off, and never force cross-platform CTAs ("watch the full version on YouTube").
- Treat 2019-era playbook advice as unverified: every platform claim must trace to a 2025/2026 signal in a reference
module or be flagged as stale.
Knowledge
Deliberately thin. This is an orchestration skill; durable declarative grounding lives in the siblings, and volatile
per-platform facts live in the reference modules, each carrying its own provenance tags. The only stable knowledge that
belongs here is the two frameworks already stated under Judgment (the seven tensions; STEPPS + Status as a Service),
because they are central, stable, and owned nowhere else.
Everything platform-specific — ranking signals, format hierarchies, thresholds, anti-patterns — is out of scope for
this file by design. Load the reference module. See references/instagram.md for the canonical provenance pattern
(PRIMARY / practitioner / internal-default, with an explicit REMOVED/CORRECTED audit trail); every reference module
should follow it.
Provenance
- Reference modules:
going_viral.tiktok, going_viral.instagram, going_viral.twitter, going_viral.linkedin —
one per platform, each with verified 2025/2026 algorithm reality, hook mappings, format winners, and anti-patterns,
provenance-tagged.
- Primary sources: Kane Kallaway (hooks, scripts, storytelling, algorithms — consumed via the sibling skills),
Devin Nash (manufactured-virality counterweight), Jonah Berger (STEPPS), Eugene Wei (Status as a Service), plus
platform-native voices per reference (Jenny Hoyos, Brendan Kane, MrBeast; Brock Johnson, Adam Mosseri, Lara Acosta;
Sahil Bloom, Naval Ravikant, Justin Welsh, Pieter Levels; Hala Taha, Richard van der Blom).
- Maintainers: the canonical research draft with full lineage and pending-transcript TODOs lives at
docs/research/youtube-library/skill-drafts/going-viral/ (not available at runtime).
