The library
for AI builders.
Two-hundred-ninety-nine triggered Claude skills — distilled from 5+ courses, production teardowns, and field experience across 60 industries. Each one quality-gated against a strict pros/cons rule, and stays out of the system prompt until its keyword fires.
From a notebook to a navigable knowledge base.
What started as 30+ skills distilled from courses is now a 300-skill library covering most of the AI-builder, operator, and founder surface area in 2026.
Three layers, one navigable surface.
Listing 300 skills flat is useless. The library is organized in three layers that work together — entry points (starter packs), workflows (recipes), and the deep content base.
Layer 1 · Starter packs
13 "start here" skills sequenced for specific roles — founder, LLM engineer, creator, AE, agency operator. Tells you which 5-7 skills to read first.
Layer 2 · Recipes
12 composition workflows — "ship first RAG in 30 days," "cut LLM bill by half in 2 weeks," "onboard first enterprise customer." Each one stitches 5-15 content skills into an outcome.
Layer 3 · Content skills
274 single-topic playbooks across industry verticals (60+), engineering patterns, GTM motions, founder operations, and compliance. Each one is a deep, opinionated read on one thing.
Layer 0 · Navigation
One meta-skill — navigate-ai-skills-library — kicks in when the right skill is ambiguous and points you at the best entry.
13 first-read paths for 13 roles.
Each starter pack is a 5-7 skill reading list that gets a specific role productive on day one. Beats reading 274 skills hoping one applies.
start-here-ai-startup-founder ·
start-here-ai-agency-operator ·
start-here-ai-startup-finance-or-coo
start-here-llm-app-engineer ·
start-here-ai-data-scientist-or-mle ·
start-here-ai-developer-relations
start-here-ai-product-manager ·
start-here-ai-customer-success
start-here-b2b-ai-sales ·
start-here-ai-marketing-lead
start-here-creator-using-ai ·
start-here-enterprise-ai-buyer ·
start-here-ai-for-healthcare-builder
12 composition workflows.
Recipes are the missing layer in most knowledge bases — they show how to combine skills for a specific outcome in a specific window. All 11 currently in the library:
recipe-ship-first-rag-app-in-30-days
For LLM engineers shipping their first production RAG app. Threads 8 content skills into a 30-day plan.
recipe-cut-llm-bill-by-half-in-2-weeks
When the API bill spikes. Caching, routing, model cascades, eval-driven downgrades.
recipe-onboard-first-enterprise-customer-in-90-days
Security review → SOC 2 → SLA → MSAs → first prod deploy. The full F500 onboarding gauntlet.
recipe-build-llm-eval-suite-from-zero
Two-week sprint from "we don't measure quality" to a versioned eval suite running on every PR.
recipe-build-ai-acceptable-use-policy-in-1-week
Internal AUP for a company adopting AI. Drafting, review cadence, edge-case carve-outs.
recipe-rollback-bad-model-release-in-24-hours
When a new model deploy regresses. Detection, rollback, customer comms, post-mortem.
recipe-prep-for-series-a-fundraise
60-90 days out. Metrics, deck, data room, intro warming, term sheet posture.
recipe-launch-on-product-hunt
The actual launch playbook — product readiness, hunter, asset pack, launch-day cadence.
recipe-respond-to-first-ai-harm-incident
When an AI product causes (or may have caused) real user harm. Triage, comms, regulatory, fixes.
recipe-migrate-llm-vendor-in-30-days
Anthropic ↔ OpenAI ↔ Gemini. Without breaking quality, latency, or the bill.
recipe-hire-first-ai-engineer-in-30-days
Sourcing, signal vs. noise on resumes, technical screen, take-home, offer.
274 deep playbooks, by domain.
The deep content base. Every skill is opinionated, source-overridden where needed, and includes failure modes. Below is the rough domain split — actual library has more granularity.
build-llm-inference-infrastructure ·
build-distributed-llm-training-pipeline ·
build-multi-model-routing-and-cascade-systems ·
deploy-llm-apps-on-serverless-gpu-platforms ·
deploy-edge-ai-and-on-device-models ·
build-streaming-llm-data-pipelines ·
build-etl-pipelines-for-ai-apps ·
build-graphrag-knowledge-bases-for-ai ·
build-cross-modal-retrieval-systems ·
fine-tune-with-rlhf-dpo-and-preference-learning ·
run-open-source-llms-self-hosted ·
build-secure-multiparty-ai-with-mpc-and-fhe ·
build-on-chain-ai-agents-and-x402 ·
build-durable-long-running-ai-agent-workflows ·
build-agent-coordination-mesh-and-swarm-systems ·
and 25 more
evaluate-llm-app-quality-with-evals-and-feedback-loops ·
add-guardrails-to-llm-applications ·
detect-and-mitigate-hallucinations ·
red-team-ai-products-systematically ·
defend-ai-apps-from-prompt-injection-and-misuse ·
secure-an-ai-app-against-data-leakage ·
design-ai-fallback-and-degradation-strategies ·
build-an-ai-incident-response-playbook ·
debug-agent-behavior-with-traces-and-replays ·
moderate-ugc-at-scale-with-ai ·
handle-ai-caused-user-harm-responsibly ·
and 14 more
write-cold-email-campaigns-with-ai ·
personalize-cold-outbound-at-scale-with-ai ·
automate-cold-email-reply-handling-with-ai ·
select-a-profitable-niche-for-b2b-outbound ·
craft-b2b-service-offers-that-close ·
position-yourself-as-a-connector-not-a-vendor ·
design-b2b-sales-cadences-that-convert ·
run-a-winning-b2b-sales-discovery-call ·
prep-b2b-discovery-calls-with-ai-assist ·
deliver-a-b2b-software-demo-that-closes ·
isolate-objections-and-close-b2b-deals-with-levers ·
analyze-sales-call-transcripts-for-coaching ·
research-accounts-at-scale-with-ai ·
rehearse-sales-calls-with-ai-roleplay-bots ·
generate-cold-call-scripts-and-objection-banks-with-ai ·
and 10 more
start-and-scale-a-niche-youtube-channel ·
maximize-youtube-cpm-and-ad-revenue ·
pass-youtube-monetization-review-fast ·
generate-viral-youtube-titles-and-thumbnails-with-ai ·
run-youtube-affiliate-marketing-for-high-revenue ·
land-creator-sponsorship-deals-with-media-kit-and-io ·
build-diversified-creator-income-streams ·
automate-a-newsletter-business ·
build-content-repurposing-pipeline ·
write-paid-social-ad-creative-with-ai ·
build-ai-powered-lead-magnet-system ·
write-micro-content-that-attracts-leads ·
run-an-ai-podcast-or-newsletter-for-authority ·
launch-and-scale-a-skool-or-circle-community ·
and 11 more
generate-ai-images-for-commercial-use ·
relight-product-photos-with-ai ·
generate-consistent-ai-characters-across-shots ·
train-a-flux-lora-on-runpod ·
run-comfyui-on-a-cloud-gpu ·
produce-ai-ugc-video-ads-ethically ·
build-multimodal-ai-products ·
build-speech-to-speech-ai-products ·
clone-a-voice-for-audio-content ·
clone-a-voice-for-ai-content ·
build-a-text-to-gif-automation-for-marketing ·
combine-ai-with-robotics-or-hardware
build-a-google-review-automation-offer-for-local-businesses ·
build-a-whatsapp-ai-agent-for-inbound-sales ·
run-instagram-sales-bot-with-dm-automation ·
sell-competitor-intel-as-a-recurring-ai-service ·
build-an-ai-booking-system-with-voice-agent ·
build-productized-ai-offers-for-ecommerce-merchants ·
build-ai-for-real-estate-agents-as-productized-offer ·
build-ai-for-insurance-agents-as-productized-offer ·
scale-an-ai-agency-past-first-clients ·
scale-ai-agency-to-holding-company ·
start-an-ai-consultancy-or-boutique-firm ·
and 9 more
raise-funding-for-an-ai-startup ·
run-an-ai-startup-as-a-solo-technical-founder ·
build-founder-to-ceo-transition-for-scaling-ai-startups ·
handle-ai-cofounder-conflict-and-equity-disputes ·
negotiate-ai-acquihire-or-talent-acquisition ·
sell-or-exit-an-ai-startup ·
start-an-ai-angel-syndicate-or-fund ·
build-an-ai-venture-studio ·
run-ai-board-and-investor-updates-monthly ·
hire-your-first-ai-engineering-team ·
handle-ai-driven-layoffs-with-integrity ·
build-ai-startup-employee-handbook-and-culture ·
and 13 more
navigate-eu-ai-act-compliance ·
handle-ai-ip-and-copyright-strategy ·
manage-ai-international-tax-and-ip-domiciliation ·
navigate-ai-tariff-and-export-control-issues ·
choose-an-open-source-ai-license ·
launch-ai-products-in-china ·
launch-ai-products-in-europe ·
launch-ai-products-in-india-and-emerging-markets ·
launch-ai-products-in-africa ·
launch-ai-products-in-latin-america ·
launch-ai-products-in-middle-east-and-gcc ·
launch-ai-products-in-australia-and-new-zealand ·
sell-ai-to-japanese-and-korean-enterprises ·
localize-ai-products-for-non-english-markets ·
and 16 more
build-an-ai-productivity-stack-for-yourself ·
kill-procrastination-with-bright-lines-and-next-action ·
prioritize-founder-tasks-with-ice-framework ·
escape-false-growth-mindset-traps-as-a-builder ·
validate-an-idea-with-physics-of-progress ·
build-in-public-as-ai-founder ·
navigate-ai-skills-library ·
master-claude-code-advanced-workflow ·
practice-vibe-coding-without-shipping-slop ·
and 3 more
Counts are approximate domain splits — the actual navigate-ai-skills-library meta-skill knows the full graph.
Three quality gates.
Every skill in the library passed the same checks. The library rule is brutal — drop weak, override source, no parroting — and applies retroactively (skills get re-audited).
Pros vs. cons
If cons outweigh pros, the skill is dropped, not patched. Better to skip a topic than ship a weak playbook.
Source override
If I disagree with the source (course, blog, book), the skill says so explicitly and provides the better take.
Distinctness check
Each new batch must be ≥3-of-5 distinct from the existing library on dimensions like industry, role, format, depth, and angle.
Citation hygiene
Fabricated papers, named-person accusations, and overstated science get caught in audit and rewritten or removed.
Want access to the library?
The full 300-skill library is private right now. I'm exploring a paid pack of starter packs + recipes, and an open-source release of the meta-skill. Follow along to find out which lands first.
Follow for updates →