- Home
- Skills
- Agents & Orchestration
- faceless video script writer
More screenshots
Works with the AI Agents you already use
faceless video script writer
now with a runnable retention checker. Scripts engineered to survive the scroll — timed beat maps, scene tables editors can shoot from, and a 24-pattern compliance gate.
$5.90
faceless video script writer
Example session with this skill installed
Write a 60-second TikTok script about the "dead internet theory" for a tech-curious audience. Use a contrarian hook and a fast 160 WPM pace.
- Read your context and instructions
- Compiled the faceless video script
I have drafted the episode contract and engineered a high-retention 9:16 script for the "Dead Internet Theory" topic, including a compliance pass for originality.
Faceless Video Script: The Dead Internet Theory
Episode Contract
- Format: 9:16 Vertical Short (approx. 55s)
- Target Viewer: Tech-curious scroll-stoppers interested in internet culture.
The ONE Takeaway: Most web traffic is non-human, but the "dead" part is about the lack of new ideas, not just bots.
- Success Metric: Retention (100%+ aim).
Retention Skeleton
| Beat | Time | Job | Re-hook / Loop |
|---|---|---|---|
| Hook | 0:00-0:03 | Contrarian Hook | Open loop: Who is talking? |
| Setup | 0:03-0:15 | The Bot Stats | [verified] traffic stats |
| Reveal | 0:15-0:30 | The AI Loop | Re-hook: The feedback loop |
| Payoff | 0:30-0:50 | The Reality | Closing loop #1 |
| CTA | 0:50-0:55 | Engagement | Follow for tech deep-dives |
Scene Table
| # | Time | Narration (words) | Dur @160 WPM | On-screen visual | B-roll keywords | Caption/overlay |
|---|---|---|---|---|---|---|
| 1 | 0:00 | The internet didn't die in two thousand and sixteen, it just stopped being for you. | 4s | Glitch effect over 2016 date | Retro web glitch | STOPPED BEING FOR YOU |
| 2 | 0:04 | Forty-seven percent of all web traffic is now bots. That is a verified fact from Imperva's latest report. | 7s | Data visualization showing 47% | Cyber security data | 47% BOTS [verified] |
| 3 | 0:11 | But the real death isn't the traffic. It is the content loop. AI is now training on AI-generated data. | 8s | Infinite mirror effect | AI feedback loop | THE CONTENT LOOP |
| 4 | 0:19 | We are reaching a point of digital inbreeding where no new ideas are actually being born. | 7s | Barren digital landscape | Empty server room | NO NEW IDEAS |
Compliance & Originality Pass
| Check | Result | Fix Applied |
|---|---|---|
| Claim Trace | Pass | All stats mapped to Imperva 2023 report. |
| Banned Patterns | Pass | No "Comment YES" or fake urgency used. |
| TTS Voice Check | Pass | "2016" spelled out as words for natural reading. |
Next steps
- Review the
[verified]status of the traffic statistics in the production notes. - Select one of the two alternate hooks provided in the
metadatafor A/B testing. - Import the scene table into your editor to align B-roll with the 160 WPM markers.
Connects securely to your tools. The creator never sees your data.
About this skill
You can spot an AI faceless script in one read: it opens with "In today's video…", explains instead of hooks, and drops a re-hook 45 seconds after the viewer already left.
Open-source faceless pipelines automate everything after the script and treat the script as a throwaway LLM call. This skill treats the script as the product.
The order of operations is inverted. Before writing a line, your agent confirms the platform's actual monetization rules (TikTok Creator Rewards needs 1min+; YouTube Shorts pays a 45% share; YouTube's "inauthentic content" policy demonetizes recycled templates) — then engineers the script against the rules that will actually pay you.
Then three hooks compete. Curiosity gap, loss aversion, contrarian claim — each written from a verifiable angle, no fabricated stats. The winner has to justify itself against the other two in one line, and you get all three for A/B testing.
What the gate catches — V2 ships scripts/retention_check.py, a zero-dependency checker that scans the finished draft for 24 failure patterns in 6 categories:
- Dead openers ("welcome back", "in today's video")
- AI-tell phrases ("delve", "game-changer", "testament to")
- Engagement bait ("comment YES") and guarantees ("guaranteed returns")
- Untagged statistics (invented precision is a compliance risk)
- TTS traps (%, $, e.g. — symbols the voice reads aloud)
- Re-hook gaps longer than your platform's cadence (25s shorts / 90s long-form)
Fix every ERROR, or the draft doesn't ship. The gate is not advisory.
Alongside the checker, every delivery includes: a timed beat map (hook → setup → reveals → payoff → CTA, with open loops tracked), a scene table pairing every narration line with on-screen visuals, B-roll search keywords, captions and WPM-accurate durations written for TTS voices, plus 2 alternate hooks and an alternate CTA for testing.
A worked example: a 60s short about Roman concrete comes back as — hook "Rome poured concrete into the sea. It got stronger." (3s, verified), a six-beat map with re-hooks at 12s and 25s, a scene table with searchable B-roll strings ("roman harbor ruins underwater"), narration at 160 WPM with every number spelled for the voice, and a compliance table showing all 8 checks passed.
Known limitations, stated honestly: the checker catches patterns, not quality — a clean pass means "no known failure modes", not "viral". The research step adds a minute before writing starts. It writes scripts; it does not render video or guarantee platform performance. Phrase-level checks are English-tuned (structural checks work in any language).
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Agents & Orchestration.
Requires user-provided sources for high-accuracy claims. Does not render video files. Logic assumes TTS voice narration at 130-165 WPM. Verification of [gap] tags is the user's responsibility.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
Trust & safety
Security scan
Security scan passed 6 days ago