/cs:webinar — Webinar & virtual-event marketing workflow. Plan a webinar from scratch (sized backward from the business goal), rescue one whose numbers disappointed (score the funnel, fix the broken stage), or turn a past webinar into an evergreen on-demand lead engine. Covers the full funnel: registration, promotion runway, show-up, live engagement, live-to-close, and segmented follow-up. Treats a webinar as a funnel, not an event.
Command: /cs:webinar [mode] [args]
The cs-webinar command is the entry point for webinar workflows: plan → promote → run → follow up, or diagnose → fix → re-run.
/cs:launch / launch-strategyemails skillplan — Design the whole motion from scratch/cs:webinar plan
Walks the intake, locks the promise + format, sizes the funnel backward from the business goal,
builds the promotion runway, and designs show-up + live-to-close + follow-up. Delivers a full plan
using marketing-skill/skills/webinar-marketing/templates/webinar-plan-template.md.
rescue — Diagnose and fix an underperforming webinar/cs:webinar rescue --input funnel.json
Scores the funnel, names the weakest stage, and returns ranked fixes targeting the actual bottleneck — not a reflexive landing-page rewrite.
evergreen — Convert a past webinar to on-demand/cs:webinar evergreen
Maps the on-demand registration → watch → follow-up automation, with honest live-vs-simulated framing.
score — Run the funnel scorer directly/cs:webinar score --input funnel.json
/cs:webinar score # embedded sample data
| Q | Asks | When | |---|---|---| | Q1 | Which mode — plan / rescue / evergreen? | Always | | Q2 | Business goal + conversion action (leads, pipeline, adoption, retention, brand)? | Always (drives the backward funnel math) | | Q3 | Audience temperature (customers / warm / owned_cold / paid_cold)? | Always (selects benchmarks) |
Read marketing-context.md first if it exists — it covers brand voice, personas, and customer language,
so you only ask for what's specific to this event.
# Mode: rescue / score — find the broken stage first
python3 marketing-skill/skills/webinar-marketing/scripts/webinar_funnel_scorer.py funnel.json
# → overall 0-100 score + per-stage rate vs. benchmark + named bottleneck
# Pipe JSON via stdin
cat funnel.json | python3 marketing-skill/skills/webinar-marketing/scripts/webinar_funnel_scorer.py -
# Demo on embedded sample data (no --help flag — run with no args)
python3 marketing-skill/skills/webinar-marketing/scripts/webinar_funnel_scorer.py
Input JSON (registrations + attended_live required; rest optional):
{
"invited": 5000, "page_visits": 1800, "registrations": 620,
"attended_live": 180, "cta_clicks": 40, "conversions": 14,
"audience": "owned_cold", "runtime_min": 45, "avg_watch_min": 26
}
Always size from the business goal backward — this stops anyone celebrating 800 registrations while 6 people buy:
Business goal: 20 sales-qualified opportunities
÷ attendee→SQO rate (~10%) → need 200 engaged attendees
÷ register→attend (~35% live) → need ~570 registrations
÷ landing-page CVR (~40%) → need ~1,425 landing-page visits
→ promotion must drive ~1,425 qualified visits
If the required visits exceed the reachable audience, fix the goal, format, or promotion budget now.
The scorer calibrates per audience temperature (warmer audiences convert better at every stage):
| Audience | Page→Reg | Reg→Attend | Attend→CTA | Attend→Convert |
|---|---|---|---|---|
| customers | 40% | 50% | 25% | 12% |
| warm | 35% | 42% | 22% | 10% |
| owned_cold | 25% | 35% | 18% | 7% |
| paid_cold | 18% | 28% | 15% | 5% |
cs-webinar-marketerwebinar-marketing/cs:aeo (get supporting content cited by AI search), launch-strategy (full launches)Version: 2.9.0 License: MIT
Copy a source-pinned command for your client. You run it yourself.
Destination: .claude/skills/cs-webinar · pinned to the source commit
# Run from your project root
git clone https://github.com/alirezarezvani/claude-skills.git .skillboard-tmp
git -C .skillboard-tmp checkout 19392f7a08264ed00486a251f5b2098321771f94
mkdir -p ".claude/skills"
cp -r ".skillboard-tmp/.gemini/skills/cs-webinar" ".claude/skills/"
rm -rf .skillboard-tmpReview the source before running. This copies files into your project; it is not a one-click install and does not verify runtime safety.
sudo apt update && sudo apt install -y gitnpm install -g @anthropic-ai/claude-code# Run from your project root
git clone https://github.com/alirezarezvani/claude-skills.git .skillboard-tmp
git -C .skillboard-tmp checkout 19392f7a08264ed00486a251f5b2098321771f94
mkdir -p ".claude/skills"
cp -r ".skillboard-tmp/.gemini/skills/cs-webinar" ".claude/skills/"
rm -rf .skillboard-tmpDestination: .claude/skills/cs-webinar
Scanner static-checks@0.1.0 · commit 19392f7a0826. Static checks cannot prove runtime safety – review the source and the exact diff before installing. How checks work.
References parent-directory traversal, which can escape a target directory.
Evidence: ../· fingerprint fa08499e14d0113b