More screenshots

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    Stage Plan: Cross-Modal Video Plans

    1

    Turns a brief plus raw material into one validated video plan: evidence ingest, a delivery promise with a motion floor, segment decomposition, reference contracts, fitted narration.

    $5

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    Stage Plan: Cross-Modal Video Plans

    Stage Plan: Cross-Modal Video Plans

    Example session with this skill installed

    Plan a 60-second launch video for our kitchen gadget. We have a 45-second hands-on demo clip, a 20-second lifestyle clip, a product photo, and a voiceover script. The real product in real hands is the whole point.

    • Read your context and instructions
    • Compiled the stage plan

    PRODUCTION PLAN — produced by the skill's protocol. Status: validated, ready for approval.

    Ingest (evidence, not assumption): demo clip probes at 44.6s, 4K, clear speech — transcribed; the transcript is the cutting map and surfaced two proof moments. Lifestyle clip: 20s, silent, soft 1080p — recorded as b-roll only, flagged risk "resolution." Product photo declared as reference R1: guide for composition, reproduce for product color; preserve/may-change axes disjoint.

    Delivery promise: footage-led (the brief's whole point is the real product in real hands), motion floor 70%.

    Segments: S1 hook 0–4s, demo's strongest moment pulled forward (footage). S2 gap 4–12s, composed problem statement. S3 core 12–38s, demo trims. S4 proof 38–50s, the demo's unscripted success moment. S5 payoff/CTA 50–60s, lifestyle clip under a composed end card. Every segment carries production method, layer, and exactly one narrative job.

    Narration: six script lines fitted to their windows at natural pace; line 4 ran 20% over its window and was shortened here, in the plan — not after synthesis.

    Validation: structural pass, zero undeclared sources. Promise check: planned motion 82% against a 70% floor — pass. Billable generation count: 0 — every beat had a sufficient cheaper source, each recorded. Plan score 11/12. Awaiting explicit approval; nothing has been generated or synthesized.

    Connects securely to your tools. The creator never sees your data.

    About this skill

    Most video projects go wrong before anything is built: segments get planned against clips nobody actually probed, references drift because nobody declared what they were for, narration is discovered to be too long only after it's synthesized, and the finished piece turns out to be a slideshow wearing a footage-led promise. This skill produces the contract that prevents all of that — a single validated plan document an executor can build from mechanically. It starts with evidence-based ingest: every supplied clip is probed (duration, resolution, frame rate, audio), read the cheapest sufficient way (transcript as a timecoded cutting map, frames for silent content), and recorded with what it's good for and what risks it carries. References are declared like contracts — reproduce, edit, or guide — with preserve and may-change axes that must never overlap. Then one delivery promise is chosen (footage-led, motion-led, composed-led, hybrid) with a motion floor attached, and the plan must prove its computed motion ratio meets that floor before approval. The timeline decomposes into segments, each declaring its production method (footage trim, billable generation, composed graphics, supplied as-is — stills never count as motion), its layer, and exactly one narrative job from a closed vocabulary. Narration is fitted to timed windows at plan time, where editing is free, instead of by repeated re-synthesis. Validation is structural plus a promise check, and the plan closes with an explicit approval step that states the generation count and cost honestly — a worked example approves a 60-second launch video with a generation count of zero, because the evidence supported cheaper sources everywhere. A scoring rubric (approve at 9+ of 12, no zero on evidence or promise fit) gives the approval decision teeth.

    • Single-axis jobs (a straight trim, one generated shot, captions only) don't need this weight of plan — the skill says so and is meant to be skipped there.
    • The plan guarantees decisions are made and validated, not that the underlying craft is good; a scored plan still needs a craft review of the finished piece.
    • Ingest quality bounds plan quality: material that can't be probed or transcribed must be flagged as a risk in the plan rather than planned around blindly.

    How to use

    1. 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. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before any payment.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    Money-back guarantee

    Doesn't do what its page promises? You get your money back.

    Trust & safety

    Security scan

    Security scan passed 24 days ago

    Permissions

    No special permissions declared or detected

    Listed24 days ago
    Updated21 days ago

    What's inside

    Frequently Asked Questions

    Because the same decision costs nothing in a plan document, a re-synthesis during production, and a re-render after assembly. The plan front-loads the cheap version of every argument — and the promise check exists because the most expensive failure, a slideshow against a footage-led promise, is invisible until the piece is finished.

    The one commitment that governs every segment choice: footage-led, motion-led, composed-led, or hybrid — with a motion floor, the minimum share of runtime that must be real motion. The plan computes its actual motion ratio and must pass its own floor before approval, not after assembly.

    Possibly, and the skill says when to skip it: single-axis jobs don't need it. The plan pays for itself when a video spans footage, generation, and graphics at once — that's where undeclared references, unfitted narration, and uncomputed promises actually bite.

    New in Image & Video