Install
$ agentstack add skill-calesthio-generative-media-skills-storyboard-previsualization ✓ scanned · ✓ verified, works with Claude Code, Cursor, and more.
Security review
✓ PassedNo issues found. Passed automated security review. · v0.1.0 How review works →
- ✓ Prompt-injection patterns
- ✓ Secret / credential exfiltration
- ✓ Dangerous shell & filesystem operations
- ✓ Untrusted network calls
- ✓ Known-malicious package signatures
What it can access
- ✓ Network access No
- ✓ Filesystem access No
- ✓ Shell / process execution No
- ✓ Environment & secrets No
- ✓ Dynamic code execution No
From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.
Verified badge
Passed review? Show it. Paste this badge into your README, it links to the public security report.
Reliability & compatibility
Declared compatibility
Compatibility is declared by the source manifest. End-to-end runtime verification is coming, see below.
We're building live execution health for every listing: tool-call success rate, median latency, uptime, and last-checked timestamps, measured, not self-reported. It isn't live yet, so we don't show numbers we can't stand behind.
How agent discovery & health will work →About
Storyboard and previsualization direction
Use this skill when a user needs to plan what a video, film, ad, animation, social cut, product piece, or generated-media sequence should look like before expensive production or generation begins. The deliverable may be text-only shot boards, illustrated boards, an animatic, 3D previs, techvis, or a generation-ready prompt plan.
Do not start by making polished images. Start by proving the sequence works: story beat, geography, framing, action, cut, duration, audio, and continuity.
Evidence classes to keep separate
- Documented facts: cite or name the source when relying on platform specs, professional tool behavior, review-tool features, or production terminology.
- Empirical observations: label anything learned from inspecting a reference, prototype, generated output, or test render.
- Production heuristics: use as judgment aids, not universal laws. Explain why the heuristic fits the current brief.
Volatile facts such as platform aspect ratios, upload lengths, ad specs, model duration limits, generation features, or pricing must be rechecked for the current project. Source examples in this skill were verified on 2026-07-10.
Decide the right previs fidelity
Match the planning artifact to the risk:
- Beat board: 6-12 key panels to test the story arc, useful for pitches and loose briefs.
- Shot board: one row per intended shot, useful before image/video generation, live action, animation, or edit assembly.
- Animatic: timed panels with scratch narration, temp music/SFX, captions, and placeholder motion to test pacing and comprehension.
- 3D previs: rough spatial scene with camera, blocking, lens/framing, and timing when depth, action choreography, or camera movement matters.
- Techvis: production-facing diagrams and measurements for camera placement, lens, crane/drone path, VFX plates, LED volume, stunt beats, or set constraints.
Documented fact: Autodesk describes previs as a way to lay out creative vision before production, using storyboards, animatics, shot lists, photos, or 3D previs scenes; it also notes that previs helps plan actor movement, camera placement, set design, lighting, and VFX integration. Toon Boom positions professional storyboarding around thumbnailing, visual storytelling, pitching boards, and timing camera moves in an animatic.
Production heuristic: Use the lowest fidelity that can answer the current question. If the open question is "does the emotional sequence work?", a beat board is enough. If the open question is "will the orbiting move reveal the product label without crossing screen direction?", use animatic or 3D previs.
Intake checklist
Before designing shots, extract or ask for:
- Intent: story outcome, audience feeling, conversion goal, or information the viewer must retain.
- Delivery: platform, aspect ratio, duration, audio/no-audio assumption, captions, language, thumbnail/poster needs, and versions.
- Inputs: script, brief, brand rules, product claims, legal disclaimers, locations, characters, assets, references, must-show moments, must-avoid moments.
- Production path: AI image generation, AI video generation, live action, animation, edit of existing footage, or hybrid.
- Review path: who approves story, style, timing, legal/brand, and final generation.
- Constraints: budget, deadline, number of generation attempts, likeness/rights limits, sensitive content, and accessibility needs.
If platform is unknown, propose a master plan plus variants. Example: a 16:9 master can be useful for YouTube/web, while a 9:16 board needs different composition, typography, and safe-zone planning for vertical social.
Build the board from story beats, not from pretty frames
Use this sequence:
- State the delivery promise in one sentence.
- Split the brief/script into beats: hook, setup, complication, proof/demo, turning point, payoff, CTA or emotional landing.
- Assign each beat a viewer question: "Where are we?", "What changed?", "Why believe this?", "What should I do?"
- Choose the minimum shot count that answers those questions with visual clarity.
- Create a continuity bible before generating individual shots.
- Only then write panel prompts, camera instructions, and review criteria.
Recommended shot-board fields:
| Field | Purpose | |---|---| | Shot ID | Stable reference for review and handoff, e.g. S03A | | Beat | Narrative purpose, not just image subject | | Duration | Initial timing estimate in seconds or frames | | Frame | Shot size, angle, subject placement, foreground/background | | Action | What changes during the shot | | Camera | Static/pan/tilt/dolly/truck/orbit/handheld/zoom plus motivation | | Blocking | Character/object positions, movement path, eyeline, screen direction | | Audio | Dialogue, VO, SFX, music cue, silence, caption need | | Continuity anchors | Character look, prop state, lighting, wardrobe, location, product orientation | | AI prompt notes | Reference inputs, shot-specific prompt, negative constraints, seed/model notes if applicable | | Handoff notes | Crew, animator, editor, VFX, or compliance instructions |
Camera, blocking, and continuity
Treat every board as a map of space and attention.
- Establish geography early when orientation matters. Use an establishing shot, insert, map-like diagram, or repeated background anchor.
- Track the line of action. Adobe and Filmmakers Academy both describe the 180-degree rule as a way to preserve spatial orientation and eyeline/screen direction; crossing it can disorient viewers unless done intentionally.
- Mark screen direction with arrows. If the subject moves left-to-right in one shot and right-to-left in the next, say whether that change is a motivated turn, a reverse angle, or a continuity problem.
- Give camera movement a job. A move can reveal information, follow action, shift power, add instability, or create scale. Avoid moving the camera merely because motion is available.
- Use shot-size changes to control emphasis: wide for geography, medium for relationship/action, close-up for decision/emotion/detail, insert for evidence.
- Reserve complex moves for moments that benefit from spatial change. If the board relies on an impossible or expensive move, flag it early and propose a simpler coverage option.
- Keep product, logo, prop, and wardrobe continuity explicit. Note which hand holds an object, whether a package is open, where a label faces, and what state a UI screen is in.
- In generated-media work, include continuity anchors in every shot prompt that needs them; do not assume the model will remember prior shots.
Production heuristic: If a viewer cannot answer "where is everyone, what changed, and what am I supposed to look at?" from the board, the shot design is not ready for production.
Reference strategy
References must serve specific decisions. Do not provide a pile of inspiration without routing each reference to its use.
Use these reference types:
- Story reference: pacing, structure, tone, or beat order.
- Composition reference: framing, negative space, foreground/background layering, lens feel.
- Motion reference: camera move, gesture, physical action, transition, or edit rhythm.
- Design reference: palette, typography, lighting, texture, set design, wardrobe, product art direction.
- Continuity reference: character sheet, product turntable, location map, prop board, UI state, wardrobe board.
- Negative reference: "avoid this kind of polish/camera/lighting/tone."
For rights and ethics, identify whether references are owned by the user, licensed, public-domain, created for the project, or used only for private analysis. Do not direct an AI system to clone a living artist, brand, copyrighted character, performer likeness, or exact frame from an unlicensed source. Translate references into observable traits such as lighting ratio, composition, camera height, color contrast, blocking, and editing rhythm.
Aspect ratio, platform, and safe-zone planning
Documented facts verified on 2026-07-10:
- TikTok Auction In-Feed Non-Spark ads list vertical 9:16 at at least 540x960 as the recommended dimension, with 16:9 and 1:1 also accepted.
- YouTube Help says Shorts uploads from desktop can be up to 3 minutes with a square or vertical aspect ratio.
- Meta/Facebook Help Center's Instagram Reels size page, visible in indexed/search snippets on 2026-07-10, reported Reels can be uploaded between 1.91:1 and 9:16, with minimum 30 FPS and minimum resolution requirements; the page may require login, so recheck from an authenticated official source for production.
Production heuristics:
- Board in the final aspect ratio whenever possible. Reframing later changes staging, text layout, and eye trace.
- For vertical social, keep the action center-biased unless the platform placement has known side UI. Leave room for captions, handles, CTAs, progress bars, and engagement buttons.
- For multi-aspect campaigns, design a "protected action box" shared across crops, then create aspect-specific boards for text, lower thirds, and product close-ups.
- Avoid making one frame carry both tiny legal text and essential action. If disclaimer text is required, allocate a calm shot or end card.
AI-generation prompt planning
Plan prompts as a production system, not as isolated magic sentences.
Create three layers:
- Continuity bible: recurring characters, wardrobe, product state, location, lighting logic, style boundaries, color palette, aspect ratio, lens language, and forbidden changes.
- Shot prompt: shot ID, narrative purpose, frame composition, subject/action, camera movement, timing, audio/silence assumption, and continuity anchors.
- Review criteria: what must be true for the output to pass, what can vary, and what requires regeneration.
Write visual prompts in concrete screen language:
- Prefer "low camera at waist height, product box in sharp foreground left, founder blurred in background right" over "cinematic product shot."
- Prefer "slow push-in over 4 seconds as the latch clicks open" over "dramatic movement."
- Prefer "same red jacket, silver zipper, coffee stain on left cuff from shot S02" over "same character."
- Use "not final text; placeholder label area only" when a model is likely to hallucinate typography.
Generation order for continuity-heavy boards:
- Create or collect approved reference assets: character sheets, product packshots, location plates, palette, typography, logo files.
- Generate key establishing frames and continuity-critical inserts.
- Generate action panels or video clips shot-by-shot.
- Assemble an animatic before polishing.
- Regenerate only the shots that fail the board criteria; do not redesign the sequence during asset cleanup unless review explicitly changes the story.
Empirical observation from current generative workflows: image/video models can produce compelling individual shots while drifting across identities, props, backgrounds, logos, text, and spatial relationships. Recent storyboard-generation research explicitly treats multi-shot continuity as a core challenge. Therefore, require a continuity bible, shot IDs, reference frames, and per-shot pass/fail criteria.
Limitations to plan around:
- Identity drift across shots, especially with secondary characters.
- Prop and wardrobe state changes after cuts.
- Unreliable exact text, UI, product labeling, hands, and small logos.
- Camera movement described in film terms may be interpreted loosely unless translated into visible motion.
- Long action chains may collapse; split them into shorter shots or key beats.
- Exact duration, frame-accurate choreography, and physical continuity may require animation, compositing, or 3D previs rather than direct text-to-video.
Animatic and timing direction
An animatic should answer timing and comprehension questions before final rendering.
Include:
- Board panels cut to estimated duration.
- Scratch VO or dialogue, even if synthetic/temp.
- Temp music and SFX only when timing depends on them.
- Captions or subtitle placeholders when the final platform will be watched silently.
- Camera moves as simple pans, zooms, parallax, or 3D camera paths only where needed.
- Slugs for shots not yet generated, marked with their shot IDs.
Review the animatic in real time. Do not evaluate only as a table. Watch for dead air, confusing geography, too many ideas in one shot, action that finishes after the cut, captions that fight the image, and missing emotional landing.
Documented fact: Toon Boom's Storyboard Pro product materials explicitly connect preproduction storyboarding with timing camera moves in an animatic. Frame.io documents single-frame, range-based, anchored, and annotated comments; Autodesk Flow Production Tracking described version comparison, frame-accurate annotations, notes, attachments, and live review in its 2026 review update.
Iteration and review workflow
Use separate review passes. Mixing every concern into one pass creates vague notes.
- Story pass: Does the viewer understand the sequence without extra explanation?
- Geography/blocking pass: Is space, eyeline, movement direction, and object state clear?
- Timing pass: Does the animatic breathe correctly? Are cuts motivated?
- Brand/legal pass: Are claims, logos, disclaimers, likenesses, and references allowed?
- Generation feasibility pass: Can the planned shots be made with available tools and budget?
- Handoff pass: Can an artist, editor, cinematographer, animator, or generation agent execute from the board without guessing?
Write notes with shot IDs and timecodes. A good note is "S04, 00:11-00:13: keep the courier moving screen-left; the reverse direction makes it look like he is returning to the store." A weak note is "this feels off."
For AI iterations, classify failures:
- Story failure: wrong beat or emotional intent. Revise board/script.
- Direction failure: prompt omitted framing, action, continuity, or duration. Revise prompt.
- Model limitation: prompt is clear but output cannot hold identity, motion, text, or physics. Change method, use references, split the shot, composite, or move to 3D/animation.
- Review ambiguity: stakeholders disagree or notes conflict. Resolve the decision before generating again.
Production handoff
A production-ready handoff should include:
- Shot board with stable shot IDs.
- Continuity bible and reference pack with provenance.
- Animatic or timing table if duration matters.
- Camera/blocking notes or overhead diagrams for spatial scenes.
- AI prompt plan with global bible, per-shot prompts, pass/fail criteria, and regeneration priority.
- Aspect-ratio and safe-zone notes.
- Audio/caption plan.
- Open questions and known risks.
- Version history: what changed and why.
For live-action or virtual production, add lens/framing intent, camera height, approximate subject distance, movement path, set/prop requirements, VFX plates, and techvis needs. For animation, add key poses, arcs, holds, transitions, expression changes, and frame-count or timing notes. For ads, add product claims, required supers, compliance timing, CTA, and platform variants.
Example: vertical product ad board and generation plan
This is an example, not a formula.
Brief: 20-second 9:16 ad for a spill-proof travel mug. Tone: practical, premium, slightly humorous. Must show leak-proof lid, one-handed opening, and CTA. Assume no live action; use AI image/video generation plus edit/composite.
Delivery promise: "Show commuters that the mug survives real chaos without making the product feel gimmicky."
Continuity bible:
- Product: matte graphite travel mug, short cylindrical body, black flip lid, small blank logo plate on front; do not invent readable brand text.
- Location: morning subway platform, cool gray tiles, yellow platform line, soft practical lights.
- Character: 30s commuter, navy coat, tan messenger bag, white earbuds.
- Style: clean premium commercial, naturalistic l
…
Source & license
This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.
- Author: calesthio
- Source: calesthio/generative-media-skills
- License: MIT
- Homepage: https://github.com/calesthio/OpenMontage
Install and usage instructions live in the source repository linked above.
Reviews
No reviews yet, be the first.
Write a review
Versions
- v0.1.0 Imported from the upstream source.