agentleFS
JuankCadavidJuankCadavid/akili-specs/.claude/skills/video-production/SKILL.md

video-production

Skill12 starsChanged 15 days agoMITView on GitHub

Trigger: video, motion video, explainer, product video, social clip, render to mp4, HyperFrames composition, Deliverable: video. Bridge between AKILI-SPECS phases and the installed HyperFrames skills: availability gate, phase map, privacy and render rules; defers every HyperFrames detail to the installed skills.

  • Reads credentials

24 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use ads-management, ai-artist and 10 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • OpenAnalystIncSkill

    ads-management

    Activate for paid advertising campaigns on Google Ads, Meta Ads, LinkedIn Ads, TikTok Ads. Includes ad copywriting, audience targeting, budget optimization, A/B testing, and ROAS tracking. Used by ads-specialist and campaign-manager agents.

    Also in 3 projects that use this

  • OpenAnalystIncSkill

    ai-artist

    Write and optimize prompts for AI-generated outcomes across text and image models. Use when crafting prompts for LLMs (Claude, GPT, Gemini), image generators (Midjourney, DALL-E, Stable Diffusion, Imagen, Flux), or video generators (Veo, Runway). Covers prompt structure, style keywords, negative prompts, chain-of-thought, few-shot examples, iterative refinement, and domain-specific patterns for marketing, code, and creative writing.

    Also in 3 projects that use this

  • truongtv22Skill

    Problem-Solving Techniques

    Apply systematic problem-solving techniques for complexity spirals (simplification cascades), innovation blocks (collision-zone thinking), recurring patterns (meta-pattern recognition), assumption constraints (inversion exercise), scale uncertainty (scale game), and dispatch when stuck. Techniques derived from Microsoft Amplifier project patterns adapted for immediate application.

    Also in 3 projects that use this

  • claude-office-skillsSkill

    email-marketing

    Email marketing automation - campaign creation, sequence building, A/B testing, deliverability optimization, and analytics

    Also in 4 projects that use this

  • sumuligeSkill

    seo-optimization

    SEO best practices for Git, documentation, and web projects

    Also in 3 projects that use this

  • mrgoonieSkill

    web-frameworks

    Build modern full-stack web applications with Next.js (App Router, Server Components, RSC, PPR, SSR, SSG, ISR), Turborepo (monorepo management, task pipelines, remote caching, parallel execution), and RemixIcon (3100+ SVG icons in outlined/filled styles). Use when creating React applications, implementing server-side rendering, setting up monorepos with multiple packages, optimizing build performance and caching strategies, adding icon libraries, managing shared dependencies, or working with TypeScript full-stack projects.

    Also in 3 projects that use this

  • +6 more from OpenAnalystInc

The file

We split this skill into the rules it gives an agent

4 rules. 0 are common, 2 are distinctive, and 2 are very specific to this project.

  • 0

    Common

    Most of the 21 edited versions we read keep these.

  • 2

    Distinctive

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Use this skill when a spec's Deliverable is video.

    line 14
  2. Distinctive:

    Only these. Everything else (flags, scaffolding, catalog, audio, captions, registry) is read from the installed skills, never recalled from …

    line 64
  3. Distinctive:

    Run the availability gate.

    line 97
  4. Very specific:

    Do the task; run check then snapshot before reporting.

    line 99

“Common” is measured: kept by at least 6 in 10 of the 21 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Video Production — AKILI Bridge

Use this skill when a spec's Deliverable is video. It decides where each video activity lives in the AKILI flow and what may run when. It teaches no composition syntax: HyperFrames' own installed skills own that, and they are always read first.

Step 0 — Availability gate (before any video work)

The gate passes when the router skill is loadable under either name: hyperframes (standalone install) or a plugin-namespaced router such as /hyperframes:hyperframes.

ResultAction
Router foundLoad it first; it picks the workflow and points to its domain skills. If the project has skills-lock.json, list the skills it lists that are not installed under the host's project skill directory, and the installed skills it does not list; report both as a warning and record it in execution.md. Never pass silently on the router alone when a lock file exists. Continue
Router missingSTOP. Print the install routes below and wait. Do not hand-write a video pipeline, do not copy HyperFrames text into the repo, do not install anything without the user's consent

Install routes (print all; the user picks):

RouteCommand / action
Host pluginthe host's HyperFrames plugin (namespaced router afterwards)
Core set, non-interactive (the route for agents)From the project root, without -g (project-scoped), with <id> from the host table: npx skills add heygen-com/hyperframes --yes --agent <id> --skill hyperframes --skill hyperframes-cli --skill hyperframes-registry --skill media-use (the eager core set, per the installed hyperframes skill's skill-lifecycle reference). Other skills are installed on demand by name
Interactive pickernpx skills add heygen-com/hyperframes — pick the core group, choose project scope
Global install (consent-gated)npx hyperframes skills update always installs globally: it writes the user's home skill directories and mirrors into other agents' global directories. Run only with the user's explicit consent

Project-scoped skill directories by host (never suggest a global directory, except the consent-gated global route above):

Host--agent idDirectory
Claude Codeclaude-code.claude/skills
OpenCodeopencode.agents/skills
Codexcodex.agents/skills
Cursorcursor.agents/skills
Antigravityantigravity.agents/skills

A project ## Skill Map row for hyperframes stays a per-developer availability note; this skill still owns the AKILI phase mapping.

Phase map

HyperFrames activityAKILI ownerRule
Intent interview, brief/akili-propose + requirements.mdThe brief is generated from approved requirements; do not re-run the interview
Storyboard plan/akili-specify design gateThe beat table in design.md is the storyboard; its approval is the storyboard approval
Optional sketch boardFirst execute taskConfirmation recorded in execution.md
check, then snapshotEach execute task's verificationContract in references/render-evidence.md
Draft render, then ffprobe/akili-testEvidence only, never the deliverable
Final look, then delivery render/akili-validateDelivery render only after the final-look approval, unless Document Control records Delivery: deferred: then the final look is on the draft render, the deferral is recorded in the validation report, and no delivery render runs
story.md (films over about 15 s)/akili-proposeThe Deliverable Outline on the Change track, the Lite proposal's Intent on the Trivial track, the Bug Diagnosis's expected behavior on the Bug track; then requirements.md. The storyline pick is that proposal's approval; no second approval
SPEC.mddesign.md (beat table and look)Generated from the approved docs and stored at the composition root; regenerated, never approved on its own. A review that changes the film changes design.md first: a beat change is a Pivot, a look change is an Adjust at the next gate
snapshot --review and its REVIEW.mdEach execute task's verificationRun as snapshot --review --describe false --output <spec>/evidence/review/<task>/ with REVIEW.md copied there; it creates no approval of its own

Requirements criteria for a video are in references/acceptance-criteria.md.

Commands this skill instructs

Only these. Everything else (flags, scaffolding, catalog, audio, captions, registry) is read from the installed skills, never recalled from here.

CommandUse
npx hyperframes checkBuild-integrity gate; runs lint first
npx hyperframes snapshot --at <t1>,<t2> --describe false --output <spec>/evidence/stills/<task>-<label>/Stills at the beat times; one output folder per run, because each run clears its folder
npx hyperframes snapshot --review --describe false --output <spec>/evidence/review/<task>/Review board for a film, with its REVIEW.md
npx hyperframes render --quality draft --output <spec>/evidence/<name>.mp4Draft render for evidence
npx hyperframes render --quality delivery --output <spec>/evidence/<name>.mp4Delivery render, after final-look approval only; not run when Delivery: deferred
npx hyperframes previewStudio preview for the user's review
npx hyperframes skills checkReads only the user's global skills directory; it never certifies a project install. Check a project install against its skills-lock.json, or reinstall with the project-scoped route; skills update is the consent-gated global route above
ffprobe -v error -show_format -show_streams <file>Output metadata

Privacy and outward actions

  1. Prefix every HyperFrames command with HYPERFRAMES_NO_TELEMETRY=1 unless the user has opted in to telemetry for this project; the skills commands are included. Also prefix check, snapshot, render, preview and init with HYPERFRAMES_SKIP_SKILLS=1: without it init checks GitHub and refreshes skills. Never put it on skills update or bare skills: the variable turns them into a silent exit 0 (the installed skill-lifecycle reference). skills check carries only the telemetry variable. Record the opt-in (or its absence) in execution.md.
  2. Consent rule. These are outward-facing: feedback (including --file-issue), publish, the hosted or cloud renders cloud, lambda, cloudrun, and snapshot --describe (reported in #6138 to send frames to a third-party model when GEMINI_API_KEY is set). Each use needs the user's explicit consent each time; earlier consent does not carry over. Without it, do not run them. The installed HyperFrames skills may suggest a post-render report; under AKILI that suggestion is off unless the consent rule is satisfied. Every snapshot passes --describe false unless the user consents to describe for that use; consent for one snapshot does not cover the next.
  3. Reports and logs written to the spec folder use relative paths; never paste home-directory paths or user names. Sanitize evidence before git add per references/render-evidence.md Evidence files.

Render rules

RuleDetail
Draft vs deliveryDraft renders feed evidence at test time. The delivery render is the deliverable and waits for the final-look approval; with Delivery: deferred it is not run and the final look uses the draft
Check is not approvalA passing check never justifies a delivery render
Quiet treeNo render, draft or delivery, while a delegated agent is active: a render is a measurement and competes for the browser, ports and build output. Render in the window after the completion report. A running tool daemon also breaks the quiet tree (a HyperFrames warm-check process was observed in the source run): wait for it to exit or stop it, and record which in execution.md
EnvironmentVideo work runs in a dedicated worktree (an Orca-managed one when Orca is present, else git worktree), one render at a time: a second render, draft or delivery, waits for the first
FontsEvery font family a composition uses is self-hosted: @font-face pointing at a local file, "bundled" families included (#6137 reports they may be fetched from Google Fonts at compile time). "Local only" is evidenced by a cold-font-cache check per references/render-evidence.md item 7; a warm-cache check is not evidence
First render on a machineBefore the first render, draft or delivery, tell the user that HyperFrames may download its own headless browser (~200 MB) into ~/.cache/hyperframes, keeps run state in ~/.hyperframes, and that rendering needs network (CDN scripts, web fonts); render only after the user consents. These home writes are the tool's runtime caches, not a skill install: the project-scoped rule governs skills only
Local onlyRenders run on this machine; hosted or remote rendering falls under the consent rule

Routing within a task

  1. Run the availability gate.
  2. Load the router; follow it to the domain skills the task touches (composition, animation, creative direction, media, captions, CLI).
  3. Do the task; run check then snapshot before reporting.
  4. Report evidence per references/render-evidence.md. A video task's PASS is never "the render finished".

Deferral table

QuestionOwner (installed skill)
Which workflow, which skill nexthyperframes
Composition rules, motion, visual direction, storyboard format, captionshyperframes (router) and its references
CLI flags, render options, troubleshootinghyperframes-cli
Sourcing images, audio, voice, media treatmentsmedia-use
Registry and cataloghyperframes-registry
Anything not listed hereThe router hyperframes decides

If an installed skill contradicts this file on a HyperFrames fact, the installed skill wins; on an AKILI rule (gates, consent, quiet tree), this file wins. Known open upstream doc conflicts (#6136 clip window, #6139 multi-mount host id): follow the router's contract and record the choice in execution.md.

Attribution

Original AKILI-SPECS skill; no HyperFrames text is included. HyperFrames is a separate project by HeyGen under Apache-2.0, installed by the user, and referenced here by name only. This skill is not a HeyGen product.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.