Diagnose and grow a founder-led YouTube channel at the channel level — "grow my YouTube channel", "why aren't my videos getting views", "help with my titles and thumbnails", "what videos should I make next", "how often should I upload", "channel strategy". Covers packaging (title + thumbnail as one unit), phase-appropriate strategy, format and series design, upload cadence vs quality, channel positioning, and analytics-driven iteration. Not for single-video hooks or scripts — those route to the hook and script sibling skills.
Installer avec Codex ou Claude Copiez ce prompt, collez-le dans Codex, Claude ou un autre assistant, puis laissez-le vérifier la page du skill et l'installer pour vous.
Une commande directe contourne le prompt de vérification. Examinez la source avant de l'exécuter.
Diagnose and grow a founder-led YouTube channel at the channel level — "grow my YouTube channel", "why aren't my videos getting views", "help with my titles and thumbnails", "what videos should I make next", "how often should I upload", "channel strategy". Covers packaging (title + thumbnail as one unit), phase-appropriate strategy, format and series design, upload cadence vs quality, channel positioning, and analytics-driven iteration. Not for single-video hooks or scripts — those route to the hook and script sibling skills.
Use this skill when the optimization unit is the channel, not the video. The sibling skills make one video better (hook, script, story); this skill decides what the channel is, which phase it is in, what its packaging discipline looks like, and which 3 videos to ship next. The two governing facts: packaging returns are non-linear (a 20% better title/thumbnail can produce 2x–100x more views — Galloway), and strategy is phase-gated (applying optimization tactics to a channel that hasn't done its reps wastes years — Galloway's three phases).
This is a strategy skill at domain altitude; the dominant verb is decide (phase classification, idea-funnel filters, packaging heuristics, positioning). It carries a secondary procedure (the channel-audit workflow) and a secondary policy layer (the phase-gate and benchmark refusals), but it is one coherent channel-level skill, not a split.
Activation
"Why aren't my videos getting views?" — diagnose a channel from described analytics or a list of recent videos.
Audit titles + thumbnails as packaging units (concepts, drafts, or published videos).
Decide what the channel is: positioning, recurring format, series design, tagline.
Choose upload cadence and the volume-vs-quality tradeoff for the channel's current phase.
Plan a founder's first videos or the next 3 videos to ship.
Run the idea funnel: turn a topic list into ranked 1-page ideas.
Decide when to broaden the channel's topic and when to stay narrow.
Do not use this skill to write or polish the first 5 seconds of one video (hook_craft_short_form), structure one script end-to-end (viral_video_script_structure), fix a flat narrative (story_driven_content_craft), choose between platforms or score a single piece for ship/kill (algorithm_aware_publishing), or set content strategy above the channel — content games, cross-platform identity (content_strategy_beyond_blogging). Full ownership map in Routing.
Judgment
Core Principles
Packaging is non-linear. Better title → higher CTR → more impressions → compounding algorithmic distribution. A 20% better title can mean 2x–100x views. Budget at least 30–40% of total effort on packaging (most creators spend ~5%). (Galloway)
The channel is phase-gated. Phase 1 = reps, Phase 2 = strategy, Phase 3 = optimization. Tactics applied out of phase are wasted or harmful. Always classify phase before recommending anything.
CCN fit caps the ceiling. A video that only works for Core viewers caps at the existing niche audience. Core + Casual + New unlocks the rest of YouTube. (Galloway)
Lessons are commodity; your relationship to the lessons is the moat. A founder channel competes on POV, not information. If the videos are interchangeable with the experts they cite, there is no channel. (Colin & Samir / Jerome)
Volume buys data, not identity. Reps deliver growth; positioning delivers a brand. 1,200 videos without a one-sentence identity is a plateau, not a strategy. (Jerome case)
Word of mouth beats algorithm hacks. If a fan can't describe the channel in one sentence over coffee, the positioning isn't tight enough. (Colin & Samir)
There is no meta. Long slow essays and fast-cut retention videos thrive simultaneously. Pick the format that fits the founder's identity, then apply CCN and packaging discipline to it. Never prescribe MrBeast-grade production to a founder. (Galloway)
Start narrow, expand outward — never reverse. Monkey-bar into adjacent niches once the narrow audience is real. Broad → narrow loses the audience you trained. (Galloway)
The Three-Phase Model (classify first, always)
Phase
Trigger
What to do
What to refuse
1 — Establishment
Fewer than 30–50 published videos
Consistency only: 1–2 videos/week, build editing/thumbnail/storytelling reps, make what calls to you
Refuse analytics deep-dives, A/B packaging programs, niche-TAM agonizing. Allow only the lightweight packaging checklist below.
2 — Improvement
30–50+ videos published consistently
Layer in strategy: idea funnel, Identity × Emotion × Action, series design, retention-curve reading
Refuse heavy per-video optimization rituals that slow cadence below weekly
Phase-gate rule: when a user in Phase 1 asks for Phase 3 work ("optimize my CTR", "deep analytics audit"), name the phase, explain the gate, deliver the phase-appropriate version instead (e.g. the packaging checklist, not an optimization program). This is the skill's most important refusal.
Packaging: Title + Thumbnail As One Unit
Title and thumbnail are one decision, evaluated together against the same idea. Run every concept through this checklist:
Thumbnail-first filter. If you cannot sketch the thumbnail before filming, the idea isn't ready. "You don't have an amazing idea then." Kill or repackage. (Galloway viral-trait 3)
Glance test. Viewers decide in under one second. Anything that requires reading, decoding, or context fails. (Galloway)
3-element rule. Maximum 1–3 visual elements (face, one prop, one short text). More = clutter = glance-test fail. (Galloway)
Exaggerate one element. Keep the design simple but exaggerate exactly one thing — contrast, scale, expression, shine. Zero exaggeration doesn't stop the scroll; two crosses into clickbait. (Galloway)
CCN title test. Would a New viewer (the "would my mom click this?" test) understand the title and want to click? Core-only titles ("Brain dump → context tree workflow in BuildOS") cap the video; universal titles ("How I run a one-person startup without losing my mind") don't.
One storyline. The title/thumbnail promises exactly one storyline — the most universal one. Multiple value-props stacked into one package means the viewer can't choose what to care about. Drip the rest inside the video. (Colin & Samir)
No misleading gap. The exaggerated element must be true to the video's payoff. Curiosity gap yes, bait-and-switch no.
Time-box. Packaging gets 1–2 hours max per video. Overthinking past that is a named small-creator failure. (Galloway)
Verdict per concept: SHIP / FIX (name the failing check) / KILL (can't be thumbnailed or Core-only with no repackage).
The Idea Funnel (what to make next)
Inputs from three sources: Internal (what worked on this channel), External (what worked for competitors / adjacent niches), Innovation (new approaches). Then five elimination filters, in order:
CCN fit?
Feasible to execute at this founder's production level?
Can you write the title + sketch the thumbnail right now?
Is the founder genuinely excited? (Test on a "YouTube council" of 2–3 peers if unsure.)
Survivors → 1-page ideas: working title, thumbnail sketch (described in words), logline, 1–3 reasons it works.
Batch ~10 one-page ideas (1–2 hours of work), gather feedback, ship the top 3–4. The 1-page idea is the cheap artifact that prevents the expensive mistake (filming a video that fails CCN).
Format, Series, and Channel Positioning
Identity × Emotion × Action worksheet (Colin & Samir) — run it at the channel level, then per video: Who is watching? What should they feel? What should they do? If the last 5 videos return scattered answers, the positioning is scattered — that's the diagnosis, not the packaging.
Decentralized sitcom. Viewers return for the personality they bonded with, not the information. Make the founder a character with stakes; make recurring collaborators/customers recurring characters.
Knock-them-off-the-list series structure. Give the audience a visible mission with a finite list (experts to learn from, problems to solve, milestones to hit) and knock one off per video. Each episode plants the next ("how do I get to X?").
Stakes before skill. Before showing the founder doing the impressive thing, seed why it matters and what's at risk — otherwise a demo is just a demo.
The tagline mechanic. Bake the one-sentence identity into a repeated phrase that closes every video. It trains viewers and word-of-mouth on what the channel stands for.
Future-headline exercise. "What headline does [relevant publication] write about this founder in 12 months?" Reverse-engineer the channel from that line.
Production floor, not ceiling (Phase 1). Phone + plug-in mic + recut method (pause silently holding eye contact, cut silences in post) + re-shoot the intro at the end when warm is enough. Candor outperforms studio polish for founder channels; escalate per-video talk-to-camera craft to viral_video_script_structure. (Oren John)
Broaden vs narrow: stay on the narrow starting niche until it's clearly won, then monkey-bar to the adjacent niche. Diagnose "started broad, now narrowing" as a positioning error to correct by re-anchoring narrow.
Procedure
Ordered channel-audit sequence. Routed steps carry a → marker and state intent only; who owns each escalated layer is declared once in Routing, never re-taught here.
Inventory. List the videos (or last ~10): titles, thumbnail descriptions, views, upload dates. Note total published count and cadence.
Classify phase from the table above. State it. Gate every later recommendation on it.
Packaging audit. Run the 8-point checklist on the 3–5 most recent packaging units. Verdict each: SHIP / FIX / KILL.
CCN audit. Tag each recent video Core-only / Core+Casual / CCN. A channel that is >80% Core-only has found its growth cap.
Identity audit. Run Identity × Emotion × Action across the recent videos; run the word-of-mouth one-sentence test; check for a tagline and a visible series mission. Scattered answers → positioning diagnosis.
Analytics read (Phase 2+ only). Apply the outlier method to the best- and worst-performing videos; check first-hour CTR habits. In Phase 1, skip and say why.
Prescribe. Cadence call for the phase, one positioning fix, and the next 3 videos as 1-page ideas from the idea funnel.
Escalate per-video work by name. → hook_craft_short_form, viral_video_script_structure, story_driven_content_craft, algorithm_aware_publishing — see Routing for which layer each owns.
Routing
Escalation is by tag, not load — name the sibling, don't inline its work. One concept, one owner; this skill retains the channel level and routes every per-video and above-channel layer.
Cadence call: uploads/week for the current phase and the production floor that makes it sustainable
Next 3 videos: 1-page idea stubs (working title, thumbnail described in words, logline, 1–3 reasons it works)
Analytics watchlist: what to measure next, each item labeled sourced or internal default — unsourced
Escalations: named sibling skills for any per-video work surfaced
Stop conditions: do not write hooks or scripts; do not produce more than 3 next-video ideas in full; do not output an analytics program for a Phase 1 channel.
Policy
Phase gate (primary refusal): refuse Phase 2/3 tactics for a Phase 1 channel (< 30–50 videos). Name the phase, deliver the phase-appropriate substitute.
Never quote absolute CTR/AVD benchmarks as fact. They are unsourced here; give channel-relative rules and label any absolute number "internal default — unsourced".
No manufactured vulnerability. Structure honest stakes into the video; never invent a sob story. (Colin & Samir)
No MrBeast-ification. Do not prescribe production upgrades, trend-chasing, or "the meta" to a founder channel; format follows identity.
No misleading packaging. Exaggerate one element, but the promise must match the payoff.
For BuildOS-owned channels, the anti-AI "lead with relief" stance applies: titles promise relief and stakes, not AI features.
Knowledge
Analytics-Driven Iteration. (This is the skill's largest declarative block and a reference-extraction candidate once youtube_channel_diagnostics is acquired — see Provenance.)
Sourced rules [practitioner — Paddy Galloway]:
First-hour (to first-24h) CTR is the only strongly predictive CTR window. Track it religiously; ignore lifetime CTR after week one — it naturally decays as impressions broaden. (Galloway) [practitioner]
Outlier comparison method. To learn from data, compare two same-niche videos with very different views and check, in order: concept novelty (borrowed from another niche?), demographics shift, device type (TV behaves differently), new-vs-returning ratio, view trajectory (spike vs build), traffic sources (search vs recommended vs external). Data narrows the hypothesis; taste completes it. [practitioner]
TAM check (4 steps): search the niche → filter out Shorts → sort by views → read the top-10 ceiling. If the ceiling is 1M+, the market exists and the constraint is execution. Use this to retire "my niche is too small." [practitioner]
Internal defaults — UNSOURCED, label them as such when used [internal-default]: absolute CTR bands (e.g. "browse CTR 4–8% is healthy") and average-view-duration bands (e.g. "≥40% AVD at midpoint") are not in this skill's source stack. If the user asks for them, give relative rules instead (compare each video against the channel's own median; investigate >2x deviations via the outlier method) and say the absolute bands are internal defaults pending sourced diagnostics. Named acquisition gap:youtube_channel_diagnostics (retention-curve reading, CTR/AVD benchmarks, A/B thumbnail testing — Tier 2 of the gap plan; Tom from Spotter / vidIQ / Creator Insider sources queued).
Examples
Task: "I'm a dev-tool founder. 14 videos in 6 months, all under 300 views except one at 2.1K. Why isn't my channel growing? Should I be optimizing my CTR?" Recent titles: "v0.4 changelog walkthrough", "Setting up our CLI with Docker", "Why we switched our queue architecture", "I quit my FAANG job to build a terminal tool (here's month 6)" (the 2.1K one), "Schema migrations deep dive".
Output (contract shape):
Phase: 1 — Establishment (14 videos < 30–50 threshold; cadence ~0.5/week, below the 1–2/week floor). Phase gate applied: declining the CTR-optimization ask — that is Phase 3 work; at 14 videos the lever is reps and packaging hygiene, not analytics. (Sourced: Galloway three-phase model.)
Packaging audit (3 most recent):
"v0.4 changelog walkthrough" — Glance test FAIL (no idea what/why), CCN: Core-only, one-storyline N/A. Verdict: KILL as packaging; the content can live inside a CCN-framed video.
"Schema migrations deep dive" — Thumbnail-first filter FAIL (nothing to sketch but a diagram), CCN: Core-only. Verdict: FIX — repackage as a story ("The database mistake that cost me a weekend").
"I quit my FAANG job to build a terminal tool (month 6)" — Glance test PASS, CCN: Core+Casual+New, one storyline, exaggerable element = founder's face + "month 6" text. Verdict: SHIP. This outlier is the channel's signal: the only CCN-fit package is the only video that traveled (internal/external/innovation funnel input: Internal).
Identity audit: Identity × Emotion × Action returns scattered answers (changelog viewers ≠ journey followers). Word-of-mouth test FAILS — no one-sentence description exists. Proposed sentence: "A FAANG engineer documenting whether a one-person dev tool can beat the giants." Tagline candidate to close every video: "Still building."
Cadence call: 1/week minimum, lower production bar (talk-to-camera + screen capture, recut method, re-shoot intro at the end) to make weekly sustainable through video 30–50.
Next 3 videos (1-page idea stubs):
"Month 7: my terminal tool made its first $1,000 (or didn't)" — thumbnail: face + phone showing a number; works because it continues the proven journey storyline (CCN pass).
"I asked 5 staff engineers to break my tool" — thumbnail: founder + one reviewer mid-grimace; knock-them-off-the-list series seed (list the 5 on screen, one per episode).
"Why I turned down a $150K consulting offer to keep building" — thumbnail: face + crossed-out number; stakes-before-skill, universal storyline.
Analytics watchlist: none yet — Phase 1. From video ~30 on: first-hour CTR per upload (sourced) and outlier comparison vs channel median (sourced). Absolute CTR/AVD bands: not provided — unsourced; pending youtube_channel_diagnostics.
Escalations: opening seconds of video #1 → hook_craft_short_form; full script for the "5 staff engineers" episode → viral_video_script_structure.
Provenance
Sources: Paddy Galloway [practitioner] — YouTube packaging strategy (CCN, TAM, phases, packaging non-linearity, idea funnel, first-hour CTR, broaden-vs-narrow) — docs/marketing/growth/research/youtube-transcripts/2026-04-29-paddy-galloway-youtube-packaging-ANALYSIS.md. Colin & Samir / Jerome (Creator Support, 1,200 videos) [practitioner] — identity vs volume, Identity × Emotion × Action, one-storyline, word-of-mouth test, tagline, series-mission, stakes — docs/marketing/growth/research/youtube-transcripts/2026-04-25-jerome-creator-support-1200-videos-ANALYSIS.md. Oren John (Art of Yapping) [practitioner] — Phase 1 production floor: recut method, re-shoot intro, candor over polish — docs/marketing/growth/research/youtube-transcripts/2026-04-27-art-of-yapping-talking-on-video-ANALYSIS.md.
Honestly thin: retention-curve diagnostics (reading drop points), absolute CTR/AVD benchmark bands, and thumbnail A/B testing procedure are not covered by the source stack. Named acquisition gap: youtube_channel_diagnostics (gap plan Tier 2 row 13 — Tom from Spotter, vidIQ/Creator Science packaging episodes, YouTube Creator Insider queued). Do not pad these areas; route or label.
Canonical origin: combo index row "YouTube channel craft for founders" in docs/research/youtube-library/skill-combo-indexes/MARKETING_AND_CONTENT.md + the three analyses above. No docs skill-draft mirror exists; this runtime file is canonical.
Maintainer: drafted 2026-06-11 under AUTHORING_GUIDE.md (first skill born under the guide). Enrich with Tier 2 diagnostics sources when acquired; re-run evals after enrichment.
