calesthio/generative-media-skills

# 3D Asset Production

This skill is for finishing an asset so another artist, engine, viewer, marketplace, configurator, game, AR app, or renderer can actually use it. The input may be generated mesh output, photogrammetry, NeRF/GSplat-derived mesh, scan cleanup, sculpt, CAD-derived polygon export, hand-modeled geometry, or a kitbashed draft. The output is a documented, validated asset package with predictable geometry, scale, materials, texture maps, LODs when needed, and provenance.

Use it when the user asks for asset prep, retopology, cleanup, optimization, baking, UVs, PBR textures, LODs, rig handoff, static mesh delivery, skeletal mesh readiness, game-ready assets, web-ready assets, GLB/glTF/USD/FBX export, DCC-to-engine delivery, or QA of a 3D model.

Do not use it for:

- Choosing or prompting a specific 3D generation provider.
- Building a whole environment, shot, level, or composed scene.
- Running motion-capture sessions or solving mocap performances.
- Exact CAD, toleranced manufacturing, BIM compliance, or simulation-grade engineering geometry.
- Character performance direction, although this skill can prepare a mesh for rigging or animation handoff.

# Neural Reality Capture

Neural reality capture is the production craft of reconstructing or rendering a real subject from many overlapping observations. In this skill, the deliverable is one of three families:

- a photogrammetric mesh with textures, usually for game, VFX, GIS, cultural heritage, ecommerce, or DCC handoff;
- a NeRF or related neural radiance field, usually for high-quality novel-view rendering from known camera poses;
- a 3D Gaussian splat scene, usually for real-time free-viewpoint playback with radiance-field appearance.

Stay provider-independent. Choose tools by evidence, constraints, and handoff needs, not by brand habit. Do not promise metrology unless the request includes surveyed control, calibrated capture, uncertainty reporting, and a qualified measurement workflow. Do not expand into text-to-3D, generative resculpting, animation rigging, material lookdev beyond captured handoff, or broad finishing work that belongs downstream.

## Evidence Labels

Use these labels in plans, reviews, and troubleshooting:

# Audio Mixing and Mastering Direction

Use this skill when audio must survive real playback: phone speakers, earbuds, laptops, TV soundbars, cinema-style trailers, podcasts, client review links, broadcast deliveries, and social feeds. Treat the mix as a production decision, not as a last-minute normalization step.

The job is to make the audience understand the foreground, feel the intended energy, avoid fatigue or distortion, and pass the declared delivery spec.

## Start with the audio contract

Before touching levels, write a short audio contract for the project:

- Delivery context: social post, YouTube upload, podcast RSS, streaming ad, broadcast, OTT, internal review, theatrical-style trailer, music video, documentary, localization, accessibility version.
- Foreground hierarchy: dialogue/VO first, performance/music first, sound-design impact first, or a changing hierarchy by scene.
- Required deliverables: full mix, dialogue stem, music stem, effects stem, M&E, narration-only, clean captions, audio-described version, alternate language mix, stereo fold-down, 5.1/Atmos printmaster, client preview.
- Target spec: use the client/platform/broadcaster spec if supplied. If no spec is supplied, choose a conservative target and label it as a heuristic.

# Captions and media accessibility direction

Treat captions and accessibility tracks as production assets, not polish. Build them from the script, audio mix, edit, localization plan, and platform delivery target. If the user asks for legal compliance, say what standards you are using and recommend review by qualified accessibility/legal counsel; do not promise legal compliance from a generated file alone.

## Start by classifying the media

Before writing or exporting captions, identify:

- Content type: prerecorded video with audio, live/near-live video, audio-only, video-only, or silent social cut.
- Audience need: captions for deaf/hard-of-hearing viewers, translated subtitles for hearing viewers, SDH, transcript, descriptive transcript, audio description, sign language, or platform-default auto captions.
- Delivery target: social burn-in, web player sidecar, broadcast/OTT, LMS/training portal, internal review, localized package, or archival master.
- Accessibility risk: essential visual text, charts, screen recordings, multiple speakers, overlapping dialogue, music/lyrics, sound-driven story beats, flashing/strobe, fast kinetic typography, low contrast, small mobile screens, or platform UI overlays.
- Final-source truth: the locked audio mix and final picture. Generated scripts, TTS drafts, and ASR transcripts are starting points, not the authority after edits.

# Dialogue editing and ADR direction

Use this skill when spoken audio must carry story, brand, instruction, character, or accessibility. Treat dialogue as a performance and an editorial contract, not just a waveform. The job is to preserve intelligibility, intent, continuity, and rights while leaving the mixer, captioner, localization team, or downstream video agent with unambiguous material.

## Evidence posture

Documented facts below come from professional standards, official platform guidance, official tool documentation, or rights organizations. Volatile platform/tool/legal facts were checked on 2026-07-10. Empirical observations are limited to listening/QC behaviors an agent should perform on the actual project; no universal audio-processing result is assumed without auditioning. Production heuristics are practical defaults to test against picture and delivery specs, not laws of audio.

## First pass: decide what kind of dialogue problem this is

Classify each line or section before touching it:

- `production-edit`: usable source audio needs cleanup, smoothing, fill, conform, or comping.
- `performance-fix`: the words are technically usable but the read, emotion, pronunciation, emphasis, or timing fails the scene.

# Localization and dubbing production

Use this skill to turn a source video into market-ready localized variants without treating translation as a word swap. Direct the localization strategy, script adaptation, voice approach, timing, accessibility, rights checks, review loop, and quality control before choosing any provider or model.

This is production guidance, not legal advice. When the work involves talent contracts, union-covered performers, biometric/voice data, likeness rights, political content, regulated claims, minors, health/finance/legal claims, or paid media in multiple jurisdictions, require qualified legal/local market review before release.

## Evidence posture

Documented facts in this skill are grounded in the sources listed at the end and were verified on 2026-07-10 unless otherwise stated. Empirical observations are recurring production patterns seen in localization/dubbing work and should be checked against the actual source media. Production heuristics are decision rules for agents; override them when a client brief, platform specification, law, accessibility requirement, or native reviewer says otherwise.

## Start by classifying the job

Identify the content type before making localization choices:

# Music supervision and scoring direction

Treat music as a production decision, not decoration. A strong music plan explains what the audience should feel, where music should enter and leave, what rights path is safe, how the cue will survive edit changes, and how the final mix will protect speech, captions, audio description, and delivery specs.

Use this skill to produce a music supervision plan, score brief, licensed-track shortlist, AI-music prompt, cue sheet draft, mix handoff, or music QA review for audiovisual work.

## Evidence labels

Use these labels when making claims or giving recommendations:

- Documented fact: supported by law, standards, official platform docs, PRO guidance, or other authoritative sources. Date volatile facts.
- Empirical observation: based on inspecting the actual edit, stems, waveform, transcript, delivery specs, or generated outputs.
- Production heuristic: a practical recommendation that often works but depends on the brief, audience, platform, and mix.

# Sound Design and Foley Direction

Use this skill to make generated media sound intentional instead of merely accompanied by music. Treat sound as story, timing, texture, space, and information. The goal is not to fill every frame with effects; it is to decide what the audience must feel, notice, understand, or believe at each moment, then design sounds that support that job.

## Evidence stance

Documented facts below are drawn from professional and standards sources verified on 2026-07-10:

- Berklee describes Foley artists as devising and recording everyday sounds for films, television, and video games, and notes that Foley work begins with a spotting session that catalogs needed effects before recording to picture.
- ITU-R BS.1770-5 specifies algorithms for measuring programme loudness and true-peak audio level; many platform specs derive from or reference ITU loudness methods.
- EBU R 128 recommends -23 LUFS for broadcast normalization in its domain, with EBU Mode momentary, short-term, and integrated metering.
- W3C WCAG guidance for prerecorded captions says captions include dialogue plus meaningful non-speech sound information.
- Section508.gov similarly says sounds needed to understand and enjoy synchronized media should be captioned, and that caption style should be consistent.
- Creative Commons licenses can permit reuse under defined conditions such as attribution, noncommercial limits, share-alike, or no-derivatives. These conditions must be checked per asset.

# Video-to-audio Foley

Use this skill to turn a silent, generated, animated, game, ad, social, or edit-ready video clip into a credible sound-effects bed. Treat video-to-audio (V2A) as a production accelerator, not as a finished mix. The job is to decide what should be automated, what must be hand-spotted, how to prompt and segment the model, how to preserve sync, and how to verify the result against the picture.

## Boundaries

Use automated V2A when:

- the clip is short enough for the available model or can be divided into coherent beats;
- the target is a believable first-pass Foley bed, social clip, animatic, ad temp track, game prototype, or atmospheric pass;
- the visible action has obvious sound causes such as water, footsteps, vehicles, impacts, machinery, crowd, fabric, animals, weather, UI gestures, or prop handling;
- the user can accept iteration and editorial repair.

Do not rely on automated V2A alone when:

# Anime and stylized 2D animation production

Anime is not a preset. It is a set of production conventions born from hand-drawn
cel animation under tight schedules: flat color, hard-edged shading, deliberately
uneven frame timing, and a shot grammar that spends motion where it matters and
freezes everywhere else. Generative tools can reproduce the *surface* of that look
easily, and the *grammar* almost never by default. The value an agent adds is the
grammar — knowing which convention a given genre and beat calls for, encoding it into
prompts and shot plans, and reviewing output against the things that actually break
the illusion.

This skill covers the craft as production knowledge. It is provider-independent. Any
model named (Sora, Kling, Vidu, Wan, PixVerse, Midjourney, etc.) is an illustrative
option, dated where its behavior is volatile — never the method itself.

# Audiobook production

An audiobook is not a long TTS clip. It is a **structured deliverable of many
chapterized files**, each of which must pass an automated loudness/noise gate,
carry the right metadata, and sound like the same performer that opened Chapter
1. The problems that dominate this work — voice drift over ten hours, a proper
noun mispronounced 40 times, a footnote that makes no sense read aloud, a file
rejected for a -58 dB noise floor, a platform that silently bans AI narration —
do not exist in short-clip synthesis. This skill is about managing hour-scale
production, not about generating one good sentence.

This is a **craft skill, provider-neutral**. Generative voice tools are
referenced by capability (pronunciation lexicon, SSML support, seed/consistency
control, per-chapter regeneration), not by brand. Specific tools and retail

# Avatar Spokesperson Production

Use this skill to plan, direct, review, and package synthetic presenter videos where a humanlike avatar or digital spokesperson speaks to the viewer. It is provider-independent: adapt the workflow to the avatar, TTS, dubbing, lip-sync, compositing, or platform tools available in the current environment.

Do not treat avatar production as only a prompt-writing task. The core production risks are identity rights, audience transparency, claims accuracy, performance credibility, brand trust, localization, and QA.

This skill is not legal advice. When rights, advertising claims, union/talent terms, regulated industries, minors, political/civic content, health/financial claims, public figures, or deceased personalities are involved, escalate to the client, counsel, brand owner, performer/voice owner, or compliance reviewer before generation or publication.

## Non-negotiable production boundaries

Separate documented facts, project requirements, and production heuristics:

- Documented fact: cite a source or project artifact.
- Project requirement: comes from the client brief, contract, brand guide, legal review, consent form, platform spec, or tool documentation.

# Brand launch film production

Use this skill to make a launch film behave like a business-critical campaign asset, not just a pretty generated video. A strong brand launch film turns positioning into a memorable audiovisual proof: the audience understands who it is for, what changed, why to believe it, and what to do next.

This is not legal advice. Treat legal, regulatory, medical, financial, securities, privacy, endorsement, talent, music, likeness, trademark, and platform-policy decisions as approval gates for the client, counsel, or the responsible compliance owner.

## Core production contract

Produce a launch film only after you can answer:

1. Who must change their mind or behavior?
2. What is the single launch promise?
3. What proof supports that promise?
4. What must not be claimed, implied, shown, copied, or simulated?

# Cinematic Trailer Production

Use this skill when the user wants a trailer-like video: a movie-style trailer, teaser, launch trailer, game trailer, documentary trailer, series promo, event reveal, high-impact product preview, or social cutdown that sells anticipation rather than fully explaining everything.

This is a production discipline skill, not a provider manual. Pair it with the relevant local pipeline, generation-provider skill, editing/composition skill, and delivery tool instructions before calling tools or exporting files.

## Core rule

Make one promise, then escalate proof of that promise.

A trailer is not a miniature full story. It is an engineered desire machine: the audience should understand the genre, stakes, tone, and reason to care while still wanting the full release. Every beat, shot, caption, sound cue, and title card should earn attention or increase anticipation.

## Start with the trailer brief

# Documentary Montage Production

Use this skill when the deliverable asks the viewer to believe, understand, or feel the significance of real people, real events, real institutions, or real-world claims. The central rule is: do not let the audience become more certain than the evidence allows.

This is not legal advice. Treat rights, defamation, privacy, platform-policy, election, medical, financial, public-safety, and vulnerable-subject issues as escalation triggers for the appropriate human editor, client approver, platform owner, or counsel.

## Non-negotiable documentary posture

Separate five things in every decision:

1. **Documented fact**: directly supported by a logged source.
2. **Corroborated interpretation**: a reasonable conclusion from multiple logged sources.
3. **Participant testimony**: attributed lived experience or opinion, not universal proof.
4. **Illustration or reenactment**: visual support that is not evidence.

# Ecommerce Product Imagery

Use this skill to produce commercially usable product imagery, not just attractive pictures. The governing rule is: the image must help a buyer understand the actual purchasable product without inventing features, quantities, sizes, colors, materials, packaging, certifications, results, endorsements, or included accessories.

## Operating principles

Separate three categories before creating anything:

1. **Documented product fact**: confirmed by physical sample, client PDP, packaging, SKU sheet, spec sheet, or approved brand guide.
2. **Creative direction**: lighting, setting, camera, styling, props, layout, crop, visual hierarchy, and mood.
3. **Regulated or platform-sensitive claim**: performance claims, health/beauty outcomes, environmental claims, comparative superiority, pricing, discounts, guarantees, safety, awards, certifications, shipping promises, and "before/after" implications.

Do not let category 2 change category 1. Do not render category 3 unless the client supplies substantiation and approves the exact wording and visual implication.

# Educational Animation Production

Use this skill to make animated instruction accurate, learnable, accessible, and production-ready. Treat the animation as an instructional product first and a media product second: every visual, motion cue, narration line, and assessment moment must support a defined learning objective.

This is provider-independent. Apply it whether the final piece is built with generated video, generated images plus motion design, vector/SVG animation, whiteboard animation, slides-to-video, Manim/D3/STEM visualization, character animation, screen capture, or a compositing runtime.

## First decisions

Before writing prompts, scenes, or code, establish the instructional contract:

1. Define the learner, not just the topic.
   - Age/grade or job role.
   - Prior knowledge and prerequisites.
   - Language proficiency and reading level.

# Explainer video production

Produce an explainer only after defining what the audience should understand, believe, decide, or do differently after watching. Treat the video as a learning and decision artifact, not as a sequence of pretty scenes.

This skill is provider-independent. Use the available generation, editing, composition, captioning, and review tools in the host environment, but keep the explainer logic, factual discipline, accessibility, and QA standards intact across providers.

## Non-negotiables

- Start from audience, prior knowledge, and a measurable learning or decision objective.
- Separate documented facts, source-backed inferences, production heuristics, and creative inventions.
- Maintain a claim log for factual, commercial, comparative, health, safety, legal, financial, scientific, and public-service claims.
- Escalate high-risk claims to the client, subject-matter expert, legal/compliance reviewer, medical reviewer, financial reviewer, or policy owner. Do not provide legal, medical, or financial advice.
- Build accessibility into script, storyboard, captions, audio, data visuals, and delivery variants. Retrofitting is weaker and more expensive.
- Re-check volatile provider/platform facts at production time: social aspect ratios, maximum lengths, safe areas, caption behavior, model inputs, model limits, pricing, regional availability, disclosure rules, and rights/licensing terms.

# Fashion Campaign Production

Use this skill to make fashion content that is beautiful, commercially useful, and honest about the garments, people, claims, and media provenance. Treat fashion production as a truth-preserving workflow: the campaign can stylize mood, setting, casting, edit, and story, but it must not invent product attributes, misrepresent fit, fake substantiation, or use a person's likeness without permission.

This is not legal advice. When a claim, disclosure, release, marketplace decision, or jurisdictional rule affects real publication, ask the client, counsel, marketplace lead, or platform specialist to approve before launch.

## Production principle

The garment is the source of truth. Everything else exists to help a customer, editor, buyer, fan, or viewer understand why that garment matters.

For every asset, distinguish:

- Documented facts: SKU, silhouette, fiber content, colorway, finish, construction, care, country of origin, size range, price, availability, tested performance, sustainability claims, endorsements, model releases, platform rules, and delivery specs.
- Creative choices: location, lighting, mood, casting archetype, pose, camera language, pacing, typography, and music.

# Food and beverage content production

Use this skill when producing food or beverage visuals for commercial, editorial, ecommerce, social, recipe, menu, packaging, or product-ad contexts. Treat the food as both a sensory subject and a regulated product: make it appetizing, but do not let aesthetics create a false product, false serving, false ingredient, false label, or unsupported claim.

This skill is not legal, nutrition, medical, or regulatory advice. For anything that could become a label claim, health claim, allergen representation, alcohol compliance issue, child-directed ad, marketplace listing, or paid media submission, prepare the creative and flag the approval required from counsel, regulatory, the brand owner, the platform, or the client.

## Operating stance

Separate facts, client-provided claims, and production heuristics.

- Documented fact: information from the client brief, product label, recipe spec, signed asset library, official regulatory source, or platform requirement.
- Client claim: a statement the client wants to communicate. Require substantiation or approval before turning it into copy, packaging text, super, caption, or alt text.
- Production heuristic: a craft technique that improves appetite appeal but must stay truthful to the product.

# Game trailer production

Use this skill to make trailers that sell the game without misrepresenting it. Treat a trailer as both creative work and regulated marketing: it must show the game clearly, preserve gameplay truth, respect platform/store rules, and survive publisher, ratings, localization, accessibility, and legal review.

This is not legal advice. Escalate to the publisher, platform representative, ratings body, or counsel when the trailer includes claims, third-party IP, paid endorsements, user data, minors, rating-sensitive content, AI-generated realism, region-specific offers, platform-exclusive language, or anything not yet implemented in the game.

## First classify the deliverable

Ask for the intended primary deliverable before structure or capture planning. If the user gives only "make a trailer," propose a primary and optional cutdowns.

- Announcement teaser: reveal existence, fantasy, tone, world, premise. Useful before gameplay is final. Do not imply final gameplay systems unless they are real or clearly labeled.
- Gameplay trailer: show actual play, verbs, UI readability, systems, player agency, combat/puzzle/exploration loop, progression, stakes.
- Launch trailer: combine fantasy, proof, availability, platforms, edition/CTA, review quotes only if approved and allowed by the storefront.
- DLC/update/season trailer: show what is new, who it is for, ownership requirements, base-game dependency, release window, and region/platform availability.

# Immersive Spatial Video Production

This skill is for passive video deliverables that surround or partially surround the viewer: monoscopic 360, stereoscopic 360, VR180, spatial video, dome/wide-FOV video, and headset-first immersive films. It helps an agent make production decisions, write capture and finishing plans, review outputs, and troubleshoot delivery failures without depending on one camera vendor, headset, or generator.

Do not activate this skill for interactive XR applications, OpenXR/WebXR app design, volumetric scene generation, Gaussian-splat/world-model navigation, realtime game-engine builds, or ordinary flat 16:9/9:16 edits unless the requested output explicitly needs spherical, stereo, spatial, wide-FOV, or headset delivery.

## Evidence Map

Documented facts below use these sources, verified 2026-07-11 unless a source date is named inline:

- Apple Human Interface Guidelines, "Immersive experiences," https://developer.apple.com/design/human-interface-guidelines/immersive-experiences. Access path: Apple Developer > Design > Human Interface Guidelines > Immersive experiences. Supports the visionOS passthrough, immersion-style, 1.5-meter boundary, movement, permission, attention-cue, and transition-comfort claims in this skill.
- Apple Human Interface Guidelines, "Playing video," https://developer.apple.com/design/human-interface-guidelines/playing-video. Access path: Apple Developer > Design > Human Interface Guidelines > Playing video. Supports the system video player, original aspect ratio, subtitles/transport controls, visionOS comfort, automatic full-immersion warning, thumbnail-track, and RealityKit closed-caption claims in this skill.
- Apple Developer Documentation, "Reading stereoscopic 3D video files," https://developer.apple.com/documentation/avfoundation/media_reading_and_writing/reading_stereoscopic_3d_video_files. Access path: Apple Developer > Documentation > AVFoundation > Media reading and writing > Reading stereoscopic 3D video files. Used only for the narrow claim that Apple documents AVFoundation workflows for reading stereoscopic video files.
- Apple Developer Documentation, "Writing spatial video," https://developer.apple.com/documentation/avfoundation/media_reading_and_writing/writing_spatial_video. Access path: Apple Developer > Documentation > AVFoundation > Media reading and writing > Writing spatial video. Used only for the narrow claim that Apple documents AVFoundation workflows for writing spatial video.

# Livestream Event Production

Use this skill when the user needs an AI agent to plan, direct, rehearse, run, troubleshoot, or hand off a live or hybrid livestream event. The event may be a webinar, conference session, town hall, launch, panel, class, fundraiser, worship service, performance, creator show, press briefing, or multi-room hybrid event.

This is a provider-independent production skill. It should produce a practical production plan, not a single-platform integration recipe.

Do not use this skill for:

- recutting, packaging, or editing an already recorded livestream after the fact;
- single-provider API integration, account setup, or dashboard-specific click paths;
- purely cinematic prerecorded video production with no live switching, live guests, ingest, monitoring, or incident plan;
- generic social short editing, podcast editing, or talking-head recut tasks unless the output is a live event plan.

## Evidence Labels

# Music Video Production

Use this skill to turn a song into a coherent, rights-aware, platform-ready video package. Treat the song as the fixed spine: the video exists to amplify its structure, artist identity, lyrics, and release goal.

Do not treat this as generic video generation. Music videos need beat maps, phrase maps, lyric accuracy, sync review, rights notes, performance direction, platform-specific cutdowns, and approval checkpoints.

## Ground rules

- Keep the master audio unchanged unless the artist, label, or client approves a radio edit, clean edit, stem edit, speed change, fade, or shortened version.
- Never invent rights clearance. Record what is cleared, unknown, platform-provided, licensed, royalty-free, public-domain, or generated.
- Do not provide legal advice. Flag clearance, likeness, endorsement, union, label, distributor, or platform-risk questions for the client, rights holder, label, manager, or counsel.
- Re-check platform rules at production time. Music libraries, AI labels, commercial disclosure tools, ad specs, and upload limits change often.
- Separate documented facts from production heuristics. Facts should trace to current sources; heuristics are creative recommendations.
- Preserve provenance: song source, lyric source, stems, edits, models, prompts, seeds when available, consent documents, licenses, platform checks, and review notes.

# Podcast production with generative tools

This skill covers producing a finished, publishable podcast episode when some or all of the
audio is machine-generated. It is provider-neutral: it names TTS engines, music libraries, and
hosts only as illustrative options, never as the method. The craft — format design, writing for
the ear, voice direction, structure, rights, loudness, delivery, disclosure, and QA — is the
same regardless of which tool renders the audio.

An episode is *done* when it (1) serves a defined show and audience, (2) sounds intentional and
consistent, (3) uses only audio you have the right to distribute, (4) meets the loudness and
file specs of its target platforms, (5) ships correct RSS metadata plus a transcript, (6)
carries any legally or platform-required AI disclosure, and (7) has passed a listen-through QA.
Skipping any of these is a defect, not a shortcut.

# Product Ad Production

Produce an advertisement that makes the right person recognize a relevant situation, understand what the product changes, believe the proof, and know what to do next. Visual polish is not a substitute for product comprehension or claim support.

This skill is provider-neutral. Select image, video, capture, voice, music, editing, and composition tools only after the advertising problem is defined. Do not let a generation model's favorite visual language determine the concept.

## Evidence language used here

- **Documented fact** means a requirement or principle stated by a regulator, standards body, or platform source.
- **Research finding** means an empirical result with a stated source and scope. It is evidence, not a guarantee for a new campaign.
- **Practitioner heuristic** means a production rule worth testing, not a universal truth.

Platform behavior and ad products change. Platform facts in this skill were verified **2026-07-09**. Re-check the exact placement specification, policy, interface preview, and destination requirements immediately before export and trafficking.

# Real Estate Content Production

Use this skill to produce real estate marketing assets that sell the property without misrepresenting it. It is provider-independent: apply it before choosing any image, video, editing, copywriting, avatar, voice, or ad platform.

This is not legal advice. Treat the rules below as production controls. Escalate uncertain legal, licensing, brokerage, MLS, or disclosure questions to the broker/client/counsel/local MLS before publication.

## Operating principle

Real estate content has a stricter truth standard than ordinary lifestyle advertising because buyers, renters, lenders, appraisers, inspectors, neighbors, sellers, tenants, MLS reviewers, and regulators may rely on it. Your job is not to create the most attractive version of a property; it is to create the most attractive truthful version.

For every asset, maintain four separations:

1. **Observed property facts**: verified from source photos, video, measurements, floor plans, survey, MLS data, inspection, seller/broker notes, or public records.
2. **Creative presentation**: color correction, composition, sequencing, music, captions, typography, and other choices that do not change the property facts.

# SaaS product demo production

Use this skill to turn a software product brief into a production-ready demo video plan, script, capture direction, generated-media prompt set, edit plan, delivery matrix, and QA pass. It is provider-independent: adapt it to real screen recording, browser automation, synthetic screens, motion graphics, avatar narration, or AI video generation without assuming a specific tool.

The core job is not "show the UI." The job is to make the right viewer understand a valuable workflow, trust the evidence, and know what to do next without exposing private data or overstating the product.

## Non-negotiables

Treat these as production constraints, not suggestions:

1. Start from audience, use case, and proof. A demo organized by the product navigation usually feels like a tour; a demo organized by the buyer's job feels like a solution.
2. Use only approved demo tenants, seeded mock data, synthetic data, or explicitly cleared customer data. Never capture live production accounts, real customer names, internal tickets, API keys, session tokens, email addresses, payment details, unreleased roadmap, admin consoles, or analytics that security/legal/client stakeholders have not approved.
3. Substantiate every objective claim before writing it into narration, captions, on-screen text, thumbnails, or ad copy. The FTC states that advertising claims should be truthful, non-deceptive, and evidence-based, and its substantiation policy expects a reasonable basis for express and implied objective claims before dissemination. Do not treat this skill as legal advice; escalate regulated, comparative, performance, financial, health, security, privacy, AI, or pricing claims to the appropriate reviewer.
4. Put required disclosures close to the claim they qualify. FTC .com Disclosures guidance emphasizes clear and conspicuous disclosure placement, proximity, prominence, duration, cadence, understandable language, and the overall net impression of the ad. If a platform or short cutdown cannot carry the disclosure clearly, change the claim or do not use that placement.

# Screen demo production

Use this skill when the central evidence is an interface being operated over time: commands run, files change, controls activate, application state changes, or a workflow completes.

This is broader than SaaS marketing. It covers instructional, support, onboarding, engineering, release-note, internal-training, terminal, IDE, browser, desktop, remote-desktop, and multi-application demonstrations.

## Evidence stance

- **Documented fact:** behavior from official capture/automation, accessibility, privacy, or security sources.
- **Production heuristic:** a practical direction choice to test with the audience and final raster.
- **Empirical observation:** a result verified in the actual product/build, rehearsal, capture, or render.

Tool and OS behavior is volatile. Facts were verified **2026-07-12**. Record product/build, browser/OS, automation/capture versions, fixtures, mocks, and permissions.

# Social Short Production

Produce an intentional short video, not merely a vertical crop. Treat each platform as a distinct playback, interface, policy, and measurement environment.

## Evidence labels

Interpret guidance by its label:

- **Documented fact** — stated by a platform, regulator, standard, or other primary authority. Recheck volatile facts before delivery.
- **Research finding** — supported in a cited study but bounded by that study's population and context.
- **Production heuristic** — a starting practice to test against the brief, audience, platform preview, and analytics; never present it as an algorithmic rule.

Do not turn a heuristic such as “hook in three seconds,” “cut every two seconds,” or “master at -14 LUFS” into a universal requirement. Platform metrics and recommendation systems change, and audience intent can outweigh tempo.

# Talking-Head Podcast Recut

Use this skill to turn existing speech-led media into truthful, useful derivative edits. The central obligation is editorial integrity: every cut must remain faithful to what the speaker meant in the source recording, and every commercial, synthetic, legal, or rights risk must be surfaced before publishing.

This is provider-independent. Use whatever local or hosted tools are available for transcription, diarization, editing, captions, audio repair, motion graphics, and export, but keep the workflow, ledgers, and review standards below intact.

## Operating principle

Treat a recut as a quotation with pictures, not as raw material to rewrite.

- Do not create a clip whose apparent claim, emotional tone, chronology, or endorsement differs from the source.
- Do not use reaction shots, cutaways, captions, titles, or music to imply agreement, conflict, surprise, confidence, causality, or product endorsement that the original source does not support.
- Do not publish synthetic voice, face, avatar, or reconstructed dialogue without explicit approval from the rights holder and the depicted/simulated person.
- Do keep enough context for the clip to be understandable without requiring the audience to know the full episode.

# UGC Ad Production

Use this skill to plan, script, generate, edit, and review short-form ads that should feel native to creator-led feeds while still being truthful, substantiated, disclosed, licensed, and ready for platform review.

This is production guidance, not legal advice. For regulated products, material performance claims, health/finance/children/political content, synthetic human likeness, or unclear disclosure duties, stop and ask the client for legal/compliance approval before producing or publishing.

## Operating principle

UGC ads work because they borrow the cues of spontaneous creator content: handheld framing, first-person speech, product proximity, lived details, casual edits, and plain-language proof. Do not let the style imply a false source, false experience, fake review, undisclosed sponsorship, or unrealistic result.

Treat every UGC ad as four linked artifacts:

1. The creative: hook, story, demo, proof, CTA.
2. The evidence file: substantiation for every objective express or implied claim.

# 2D character rig animation

Use this skill after the character design and approved model/reference sheets exist. Its job is to turn that design into a reusable performance system whose parts stay connected, poses remain on-model, facial states read clearly, and animation can be reproduced across shots and runtimes.

The rig may be SVG, layered vector art, raster cutouts, a bone-and-mesh puppet, or a hybrid. Runtime APIs differ; the production contract does not.

## Boundaries

This skill owns:

- part decomposition, draw order, hierarchy, pivots, controls, and constraints;
- bind/rest pose and coordinate conventions;
- rigid cutout versus deforming mesh decisions;
- reusable body poses, facial states, visemes, gestures, and cycles;

# Character Design Continuity

Use this skill to make a character stay recognizably the same while still acting, aging, changing costume, moving through shots, or adapting to a campaign's style. Treat continuity as a production system, not a single prompt trick.

## Ground rules

Separate three kinds of guidance in your reasoning:

- **Documented facts**: cite a source or the user's approved materials.
- **Empirical observations**: describe what happened in the latest generation or test batch.
- **Production heuristics**: useful craft rules that improve odds but are not guarantees.

Do not promise perfect identity preservation from text prompts alone. Current image and video systems can drift on facial structure, body proportions, hands, age, wardrobe, markings, props, and style. Strong continuity comes from a character bible, reference strategy, constrained prompt planning, iterative selection, and QA.

# Direct cinematic shots as editorial decisions

Treat a shot as a controlled change in audience knowledge, attention, and feeling—not as a bundle of cinematic adjectives. Decide what the audience must perceive, choose the camera relationship that makes that perception likely, and specify only the variables needed to preserve it through production and editing.

Use three evidence labels throughout the work:

- **Established fact**: optics, geometry, or supported perceptual evidence.
- **Production convention**: a broadly used continuity or craft practice that may be broken deliberately.
- **Production heuristic**: a defensible starting point, not a law or guaranteed emotional effect.

Do not claim that a high angle always means weakness, a low angle always means power, a wide lens always feels energetic, or handheld always feels real. These are contextual readings. State the intended effect and test whether the complete image supports it.

## Begin with the dramatic job

# Lighting direction for generated media

Use lighting as story structure, not decoration. State what the audience should understand from the light, then specify the visible lighting behavior that an image or video model can render: source motivation, direction, quality, color, contrast, exposure, falloff, catchlights, separation, practicals, and continuity.

This skill is provider-independent. Translate the guidance into the syntax of the chosen image, video, avatar, or composition tool after reading that tool's own instructions.

## Evidence labels

Use these labels in plans, prompts, and reviews when claims matter:

- **Documented fact**: supported by a cited lighting, cinematography, color, accessibility, or manufacturer source.
- **Empirical observation**: based on direct inspection of generated outputs or references in the current project.
- **Production heuristic**: a repeatable craft rule that often works but should be tested against the brief and model behavior.

# Performance direction for generated media

Use performance direction to make a generated person or character appear to want something, react to pressure, and communicate through voice, face, gaze, and body. Do not ask a model for a generic emotion and hope it supplies acting. Give playable circumstances, beat-level actions, physical behavior, vocal behavior, continuity constraints, and a way to judge the take.

## Evidence stance

Documented facts used by this skill:

- Stanislavski-derived actor training commonly analyzes a role through given circumstances, tasks/objectives, actions, beats, obstacles, and physical action; for AI use, translate these into concise, playable instructions rather than academic labels. Source cross-check: [Stanislavski system overview](https://en.wikipedia.org/wiki/Stanislavski%27s_system) and [NYFA summary of Stanislavski questions](https://www.nyfa.edu/student-resources/stanislavski-in-7-steps-better-understanding-stanisklavskis-7-questions/).
- Lip-sync systems often map speech sound to visemes, not one mouth shape per phoneme. Microsoft documents visemes as visual phoneme descriptions and notes many phonemes can share one viseme; Meta's Oculus Lipsync documentation describes interpolation between visemes over time. Verified 2026-07-10: [Microsoft viseme docs](https://learn.microsoft.com/en-us/azure/ai-services/speech-service/how-to-speech-synthesis-viseme), [Meta viseme reference](https://developers.meta.com/horizon/documentation/unreal/audio-ovrlipsync-viseme-reference/).
- Professional dubbing guidance treats timing, natural language, audio level, and physical mouth/gesture sync as production constraints. Verified 2026-07-10: [Netflix films/series dubbing guidelines](https://partnerhelp.netflixstudios.com/hc/en-us/articles/11220839130771-Dubbing-Creative-Guidelines-Films-Series-English), [Netflix nonfiction simil-sync guidance](https://partnerhelp.netflixstudios.com/hc/en-us/articles/11220999686163-Dubbing-Creative-Guidelines-Nonfiction-Unscripted-English).
- Current avatar tools expose different levels of performance control: some require clean training footage and consent capture, some accept a script and avatar selection, and some offer sentiment/emotional-state controls. Verified 2026-07-10: [Synthesia Studio Avatar requirements](https://docs.synthesia.io/docs/studio-avatars), [D-ID quickstart](https://docs.d-id.com/docs/quickstart), [HeyGen Digital Twin docs](https://developers.heygen.com/generate-avatar-video).
- Current AI video prompting can include character references, longer clips, extension, and batch workflows in some providers; treat capability limits as provider- and date-specific. Verified 2026-07-10: [OpenAI Sora 2 prompting guide](https://developers.openai.com/cookbook/examples/sora/sora2_prompting_guide).
- Performer likeness, voice, and digital replica use raise consent, compensation, use-limit, storage, and disclosure duties. Verified 2026-07-10: [SAG-AFTRA AI resource page](https://www.sagaftra.org/contracts-industry-resources/member-resources/artificial-intelligence), [SAG-AFTRA Digital Replicas PDF](https://www.sagaftra.org/sites/default/files/sa_documents/DigitalReplicas.pdf), [NAVA synthetic voice guidance](https://navavoices.org/synth-ai-info/).

# Precise video description

Use this skill when an existing video must be represented faithfully in production language. The output should let another agent or reviewer reconstruct what is visibly present, where it is, how it changes, and how the camera records it without inventing intent or atmosphere.

This skill describes observed media. It does not direct a future shot, infer a filmmaker's psychology, write accessibility captions, or execute a provider API.

## Evidence stance

- **Observed fact:** directly visible in reviewed frames or playback.
- **Measured fact:** derived from file metadata or a reproducible tool, with method and tolerance.
- **Inference:** plausible but not visually established; label it or omit it.
- **Production heuristic:** a practical description rule, not a claim about the source.

The five-aspect structure below is informed by Lin et al.'s CHAI video-language specification and professional cinematography practice. Sources were checked **2026-07-14**. CHAI's complete CameraBench-Pro primitive catalog was not fully available under a stable public package during authoring; do not reconstruct or copy an unavailable taxonomy from screenshots.

# Production design direction for generated media

Use this skill to act as a production designer for generated images, videos, campaigns, and mixed-media sequences. Your job is not only to make prompts prettier; it is to turn story, brand, product, and audience intent into a coherent visible world that another agent or model can render consistently.

Production design direction covers:

- spaces: sets, locations, geography, architecture, room logic, negative space;
- objects: hero props, set dressing, packaging, signage, furniture, tools, wear, evidence of use;
- visual systems: color, value, material, texture, scale, pattern, typography, graphics, UI-in-environment;
- world rules: era, culture, class, climate, technology level, maintenance, history, constraints;
- continuity: what must repeat, what may evolve, and what must never appear;
- generation planning: reference use, prompt decomposition, shot-by-shot design notes, QA, and repair.

## Ground truth to keep separate

# 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.

# Direct the visual system

Create a visual grammar that another production agent can execute repeatedly, not a pile of adjectives and not a clone of a reference. Translate intent into observable rules, encode the rules in portable artifacts, prove them on representative frames or shots, and repair the system rather than patching isolated outputs.

## Evidence labels used here

- **Documented fact** — a statement supported by a standard, first-party technical document, or institutional source.
- **Research finding** — a result from a cited empirical study; apply only within the study's limits.
- **Production heuristic** — a practical rule of thumb. Treat it as a testable default, not a law.

Keep these categories distinct in the direction packet. Never present taste, convention, or a provider behavior as a standard.

## Deliverable contract

# Asset continuity management

Use this skill to keep generated-media projects reproducible, reviewable, and visually/aurally consistent across many assets, revisions, tools, and deliverables. Treat continuity as a production control system: every asset gets an ID, every change becomes an event, every derivative points back to its sources, and every continuity-critical choice is captured before more generation compounds the mistake.

This skill is provider-independent. Do not assume any particular generator exposes seeds, request IDs, input hashes, content credentials, prompt logs, or model versions. Capture them when available; explicitly record `unknown`, `not_provided`, or `not_applicable` when they are not.

## Core rule

Never manage continuity by memory, filenames alone, or chat history alone. Maintain a project-local continuity package with three living records:

1. `asset_manifest`: one row/object per file or generated unit.
2. `source_reference_ledger`: one row/object per source, reference, prompt, seed, model/tool context, approval, and rights/usage constraint.
3. `continuity_bible`: canonical descriptions and locked references for characters, products, locations, wardrobe, brand, voice, music, captions, and recurring design motifs.

# Generated Media QA

Treat QA as a release decision, not a vibe check. Judge the deliverable against the approved brief, platform specifications, legal/safety constraints, and the audience context. Record enough evidence that another agent or producer can reproduce the decision.

## Keep three evidence lanes separate

Documented facts are requirements from the brief, platform specs, legal/policy guidance, delivery standards, accessibility standards, or provider documentation. Cite them or name the source and verification date.

Empirical observations are what you directly measured or inspected in the asset: frame size, duration, loudness, sync offset, OCR output, transcript mismatch, visual artifact, metadata, or a timestamped defect.

Production heuristics are professional judgments used when no explicit spec exists: whether a hand artifact is audience-visible, whether a product packshot feels trustworthy, whether an accent is intelligible for the target market, whether a social caption is too fast for mobile. Label these as heuristics and avoid pretending they are universal standards.

## Intake before review

# Media provenance and rights governance

Use this skill to make generated media releasable, auditable, and explainable. It is not legal advice. It is a production governance workflow that helps an agent identify rights evidence, missing permissions, disclosure duties, and escalation points before a client or platform receives the asset.

Default posture: do not claim an asset is "cleared," "copyright-safe," "commercially safe," or "owned" unless the production record shows the exact basis for that statement. Prefer "approved for this release under the documented assumptions" plus a dated caveat.

## Separate facts, observations, and heuristics

Treat these as different evidence classes in your work product:

- Documented facts: directly supported by contracts, licenses, provider terms, platform rules, consent forms, official guidance, or technical metadata. Cite the source and verification date for volatile facts.
- Empirical observations: things observed in the actual workflow, such as "metadata survived this export" or "the uploaded platform copy retained no visible credential." Record the test, file hash, date, tool, and result.
- Production heuristics: conservative operating rules used to reduce risk when legal certainty is unavailable. Label them as heuristics, not law.

# Media QC and delivery

Use this skill at the finish line of media production: after creative approval or near-final assembly, before the files are sent to a client, platform, publisher, broadcaster, streamer, localization vendor, or ad operations team.

Do not invent universal export specs. Delivery is governed in this order:

1. The signed client/platform/broadcaster delivery specification.
2. The campaign media plan, insertion order, distribution platform, or localization brief.
3. Official public platform specs verified for the current date.
4. A conservative house mezzanine plus platform-ready derivatives, clearly labeled as a production heuristic.

If the destination is unknown, ask for it. If the user needs a fast default, deliver a high-quality mezzanine plus common web/social derivatives and state that the package is provisional until the destination spec is confirmed.

## Evidence labels to use in every QC decision

# Reference Media Analysis

Use reference media to understand what works, not to reproduce protected expression, private identity, or misleading claims. Treat each reference as evidence for production decisions: extract transferable intent, structure, craft, constraints, and quality bars; avoid copying distinctive expression unless the client has explicit rights and approvals for that exact use.

This skill is not legal advice. It gives production triage, documentation, and escalation triggers. When rights, likeness, voice, trademark, privacy, endorsement, or jurisdictional questions materially affect release, pause and ask for client/legal approval.

## The governing rule: extract function, not identity

For every reference, separate:

- Documented facts: what the file visibly or audibly contains; metadata; rights and consent statements supplied by the client.
- Production inferences: why the reference works; what creative function each element serves.
- Heuristics: safe ways to adapt the function into a new work.
- Prohibited copying risks: distinctive expression, private identity, voice/likeness, trademarks, product claims, or platform-regulated synthetic media disclosures.

# Video description oversight

Use this skill when language about a video must be accepted as evidence-quality metadata rather than plausible prose. It governs a correction loop:

```text
source video + specification + pre-caption
  -> evidence-backed critique
  -> revised post-caption
  -> independent acceptance review
  -> versioned triplet and decision record
```

The reviewer verifies what language says about media. Technical file integrity, visual artifact QA, accessibility captions, creative intent, and model post-training belong elsewhere.

# Color grading and finishing direction

Use this skill to guide the color and finishing stage of a video project, whether the material is fully AI-generated, camera-originated, motion graphics, screen capture, avatar footage, product renders, or a mixed-source edit.

Treat color as both a technical pipeline and a storytelling surface. A good finish makes shots belong together, preserves required colors, supports the intended emotion, survives compression and platform conversion, and gives downstream editors/compositors unambiguous instructions.

## Start by separating the job

Before proposing fixes, classify the work into three layers:

- **Color correction:** normalize the image so shots are usable and consistent. Balance exposure, white balance, black/white points, contrast, saturation, and obvious color casts. Documented fact: Blackmagic describes balancing color as the first correction step, using primary tools to adjust brightness and levels and remove unwanted tints before creative grading begins ([Blackmagic Design Resolve Color page](https://www.blackmagicdesign.com/products/davinciresolve/color), verified 2026-07-10).
- **Color grading:** create the intended look after the material is technically balanced. Shape mood, palette, contrast style, warmth/coolness, density, texture, highlight behavior, and scene-to-scene emotional progression.
- **Finishing:** prepare the final master for review and delivery. Check legal/video levels, gamut, color tags, SDR/HDR target, accessibility contrast, caption/graphic readability, compression resilience, safe areas, artifact repair, and handoff materials.

# Editing and montage direction

Use this skill when the edit itself determines whether a video works. Your job is to make decisions about sequence, duration, shot function, rhythm, continuity, audio/visual relationship, platform fit, and revision logic. Do not treat editing as "put the clips in order." Treat it as the craft of controlling what the viewer notices, understands, feels, and remembers.

Work provider-independently. The source clips may be human-shot footage, generated clips, stills animated in a composition engine, screen recordings, avatar footage, stock, captions, product UI, diagrams, or music. The same edit principles apply, but generated media needs stricter provenance, continuity, and QA because shots may not share stable identity, lighting, geometry, text, or physics.

## Evidence classes to keep separate

Documented facts:

- Film-analysis sources commonly describe editing as the relationship between shots and the construction of graphic, rhythmic, spatial, and temporal relationships. Source: Yale Film Analysis, "Part 4: Editing" (<https://filmanalysis.yale.edu/editing/>).
- J-cuts introduce the next shot or scene's audio before the visual cut; L-cuts let prior audio continue after the visual cut. Source: DINFOS Pavilion, "How to Edit Video with the J-Cut and L-Cut," last verified 2025-01-16 (<https://pavilion.dinfos.edu/How-To/Article/2286253/how-to-edit-video-with-the-j-cut-and-l-cut/>).
- Prerecorded synchronized media with audio needs captions under WCAG 2.2 Success Criterion 1.2.2 unless the media is an explicitly labeled media alternative for text. Source: W3C WAI (<https://www.w3.org/WAI/WCAG22/Understanding/captions-prerecorded.html>).
- Practical caption placement guidance includes safe-zone placement and no more than two lines per caption. Source: DCMP Captioning Tip Sheet (<https://dcmp.org/learn/225-captioning-tip-sheet>).

# Motion Graphics Direction

Use this skill to turn a video brief into motion graphics that explain, emphasize, brand, or pace the piece. Treat motion as editorial direction: every move should reveal meaning, guide attention, clarify structure, or create a deliberate feeling.

This is provider-independent. Use it before choosing a model, renderer, animation library, editing tool, or compositing environment.

## Evidence stance

Documented facts:

- WCAG 2.2 requires web content to avoid flashes above the three-flashes threshold and documents that interaction-triggered motion should be disableable unless essential. Verified 2026-07-10 from W3C WCAG Understanding pages.
- W3C Technique C39 documents `prefers-reduced-motion` as a way to let users prevent animations. Verified 2026-07-10.
- WCAG contrast guidance sets 4.5:1 for normal text and 3:1 for large text at Level AA, with higher AAA ratios. Verified 2026-07-10.
- Apple, Material Design, Microsoft Fluent, and IBM Carbon all treat motion as a design-system behavior with purpose, timing, easing, accessibility, and consistency constraints rather than decoration. Verified 2026-07-10.

# Still Image Retouching and Finishing

This skill is for finishing existing still images. It assumes there is a captured, scanned, rendered, or composited source image and that the agent's job is to make production decisions about development, retouching, color, metadata, output, and review. It is tool-agnostic: Lightroom, Camera Raw, Capture One, Darktable, RawTherapee, Affinity Photo, Photoshop, GIMP, command-line processors, DAM systems, RIP software, and lab portals are possible instruments, but the skill's decisions are about craft and delivery rather than any one vendor.

Use it for:

- RAW intake, rendered-file intake, selects, versioning, sidecars, and handoff notes.
- Nondestructive development: exposure, white balance, tone mapping, lens corrections, local masks, snapshots, proof copies, layers, smart objects, adjustment stacks, and reversible repair.
- Color-managed finishing: working spaces, embedded profiles, monitor profiling, ICC soft proofing, rendering intents, output profiles, print proofing, and web/social conversions.
- Retouching: dust, sensor spots, blemishes, flyaways, fabric wrinkles, product cleanup, architecture perspective, food freshness, documentary limits, compositing cleanup, and client/legal review.
- Generative fill or AI repair decisions only as post-production interventions on existing stills, including disclosure, provenance, and rights notes.
- Sharpening, noise reduction, resampling, crop variants, batch consistency, metadata, accessibility metadata, export recipes, QA, and archive/delivery packages.

Do not use it for:

# Title and kinetic typography direction

Use this skill to make text behave like a designed scene, not a decorative overlay. Treat every text element as a timed editorial decision with a job: identify, explain, sell, signal emotion, guide attention, or make speech readable.

This is provider-independent. It applies whether the final asset is built in After Effects, HTML/CSS/GSAP, Remotion, Lottie, a game engine, subtitles, an AI video prompt, or a still-image/video generator. When exact text must remain correct, prefer editable/rendered typography layers over asking an image or video generator to invent letterforms.

## Separate facts, observations, and heuristics

Documented facts:

- WCAG 2.2 requires web content to avoid more than three flashes in any one-second period unless flashes are below threshold; apply this as a conservative safety rule for video text even when the deliverable is not a web page.
- WCAG guidance on moving, blinking, scrolling, or auto-updating content lasting more than five seconds supports giving viewers control where the medium allows it; when control is impossible, reduce or simplify nonessential motion.
- Unicode and W3C internationalization resources define line breaking as script-specific; do not assume English word-wrap behavior works for CJK, Thai, Arabic, Indic scripts, or mixed-script titles.
- TikTok ad safe-zone documentation says key elements such as text and logos must be inside the relevant safe zone, and that safe-zone size changes with ad dimension, caption length, and add-ons. Verified 2026-07-10.

# VFX compositing direction

Use this skill to make composite shots feel photographed together, not stacked together. Treat compositing as shot integration across color, motion, geometry, light, lens, material contact, atmosphere, artifact repair, and delivery handoff.

Do not use this skill as a substitute for legal clearance, colorist judgement, an on-set VFX supervisor, or a compositor operating a specific node graph. Use it to brief, plan, supervise, and review provider-independent compositing work.

## Evidence posture

Documented facts below come from professional documentation and standards sources listed at the end. Tool-specific documentation facts were verified on 2026-07-10 and may change.

Empirical observations are repeatable checks an agent can perform by reviewing frames, scopes, mattes, tracks, and before/after renders.

Production heuristics are practical rules of thumb. They are not universal laws; override them when shot evidence, client direction, or a facility spec requires a different approach.

# Audio-reactive video composition

Use this skill to turn an approved audio source into an auditable cue map and then into bounded visual behavior. The central contract is:

$$
\text{source audio} \rightarrow \text{measured features} \rightarrow
\text{confidence-bearing cues} \rightarrow \text{reviewed anchors} \rightarrow
\text{deterministic visual timeline}
$$

Do not equate a detector output with editorial meaning. An onset is not automatically a beat, cut, drop, or lyric accent; an acoustic cluster is not automatically a verse or chorus.

## Evidence stance