Borrowing it
Nothing to install: this file belongs to ucsandman/marketing-studio. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/ucsandman/marketing-studio/main/CLAUDE.mdgit clone --depth 1 https://github.com/ucsandman/marketing-studioWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/instructions/ucsandman/marketing-studio/claude-md)<a href="https://agentmods.dev/instructions/ucsandman/marketing-studio/claude-md"><img src="https://agentmods.dev/badge/instructions/ucsandman/marketing-studio/claude-md.svg" alt="Measured on agentmods" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.01397 | $0.01397 |
| Opus 5 | $0.00698 | $0.00698 |
| Sonnet 5 | $0.00279 | $0.00279 |
| Haiku 4.5 | $0.00140 | $0.00140 |
Grade A, and why
marketing-studio CLAUDE.md scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured yesterday.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 77 lines — stays where its author put it; the contents beside it link to each section on GitHub.
animations: agent-driven marketing studio
Remotion (studio/) renders all final video for ALL products/brands. Production inputs,
evidence, renders, and delivery assets live in the owning product repo under
marketing/assets/<brand>/; the engine clone is not a production destination.
brands/.json holds per-product tokens
(zod-validated via studio/src/lib/brand.ts; mark components in studio/src/brands/marks.ts);
templates resolve getBrand(brandId) and never hardcode brand values. Feeders:
feeders/blender (headless bpy), feeders/capture (Playwright), feeders/comfy (ComfyUI,
non-load-bearing). Spec: docs/superpowers/specs/2026-07-09-animation-studio-design.md.
skills/ is the source of every skill this repo ships (installed to ~/.claude/skills via
scripts/install-skills.mjs; --check reports drift). /launch is the gated launch CLI
skill (skills/launch, tested by launch/tests/skill.test.ts); /announce is the studio's
browser-driven announce flow; /ship-it chains /marketing then /launch.
launch/ is the distribution layer, folded in from the launch-engine repo on 2026-09-02: its own package (Node 24, vitest, tsc, vite dashboard), its own conventions in launch/CLAUDE.md. Posting is dry-run unless --live; never make a test hit a live provider. Social keys live in launch/.env; infra credentials are MCP-only.
Read docs/PLAYBOOK.md before any asset or feeder work (short: engine map, onboarding, process, delivery contract) plus the docs/playbook/.md for the feeder or stage you touch — the topic index is in PLAYBOOK.md. Do not read topic files for feeders you are not using. User-level skills (/logo-reveal, /social-clip, /product-demo, /launch-video, /og-assets, /audio-track, /marketing) drive this repo from any repo.
Rules:
- Brand color rules live in each brand's JSON
voice(noban: profit = gold #d6c23c NEVER green; green = safe/simulation only). - Rendered proof: visual work is not done until a rendered frame was inspected; final assets are not done until the user saw them.
- Smoke check before claiming done: node scripts/smoke.mjs (every composition listed).
- A film is not done without audio, and audio means a voice explaining the product.
Every delivered mp4 (launch lock, every
*-final*.mp4, every postkit file that is not a-silentvariant, the newestfilm-vN) carries a mastered soundtrack: music bed + narration + SFX cues, at master-audio's TARGET_I. Music-only is a recorded, deliberate exception (--music-only), never a default; "none" is not an intake option.node scripts/check-audio.mjs <brand>is a HARD gate (exit 1) and runs before any delivery claim. Two silent postflop cuts shipped on 2026-09-01 because nothing enforced this; do not make it three. - Bespoke films (
studio/src/films/<brand>/, e.g. PostflopFilm) are scored withscripts/build-<brand>-film-audio.mjs(copy + cue source of truth, generated manifest props/-film-audio.json) thenscripts/score-film.mjs <brand> <film.mp4>; template films keep build--audio.mjs + audio-track. - Seven mechanical gates run per brand: six judges (judge-av-sync, judge-demo-pacing,
judge-palette, judge-motion, judge-drift, check-budgets) plus check-audio. The
judges are advisors: exit 0 unless --strict; check-budgets and check-audio are hard.
judge-drift runs LAST — it scores the product's
marketing/assets/<brand>/as a SET, so running it before everything has rendered scores an incomplete set and the number means nothing. It has no absolute threshold on purpose (nobody publishes one); calibrate with--ref <dir>against approved assets, never by inventing a constant. - Generated props JSON is edited only via its builder script (scripts/build-*-props.mjs).
- Asset copy traces to
<product>/marketing/assets/<brand>/marketing/brief.json(agent-synthesized, gated by scripts/lint-copy.mjs and the storyboard approval); builders overlay brief copy — never hand-edit copy into generated props. - Engine-local out/, assets/, props/, examples/, and studio/public/*/ are gitignored build products, never production workspaces.
- Blender via BLENDER_PATH in .env; ComfyUI on :8000/:8188 with documented fallback.
- Unreal via UNREAL_PATH in .env (feeders/unreal, docs/playbook/unreal.md); plates are PNG sequences staged like Blender output.
- A storyboard or any approval artifact is handed over annotated, never as bare frames: shot number, timing, ground, the camera move drawn on the frame, the narration line under it, and one line saying what the whole thing is. Render the page and read it as a stranger before handing it over (2026-09-03: an unlabeled tile of five Unreal frames was unreadable to Wes). In a terminal session, open it locally.
- The viewer test comes before the gates. A film is not done until the agent has WATCHED it with sound at full length and written a viewer's verdict ("would I share this?"); the seven gates measure properties, not quality, and on 2026-09-03 all of them passed a film Wes called bad on every axis. The verdict happens at the animatic (half-scale plates + scratch VO), the cheapest point to stop. The DashClaw NIGHT SHIFT film is PARKED: do not iterate on it; a new attempt starts from a script read aloud and an animatic Wes reacts to in the first hour, with real assets and a chosen voice.
- Every gate ends with a retro (approval, ship, wrap, any correction): what worked, what did not, the one change, written to docs/ERRORS.md or PLAYBOOK in the same turn. The next film starts by reading the last film's retro.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- yesterday Changed · +1 lines · +6 tokens per session 5ad86c55b4a6
- 2d ago Changed · +3 lines · +46 tokens per session 171ec0f9d173
- 3d ago Changed · +37 lines · +765 tokens per session 00fcd344314f
- 8d ago First seen · 36 lines · 580 tokens per session scan A 0bdb90ea189a
marketing-studio CLAUDE.md is an instructions file published in the GitHub repository ucsandman/marketing-studio (227 stars, last pushed 2d ago), licensed MIT. It adds 1,397 tokens to every session, about $0.0070 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other instructions, from other repositories
next.js AGENTS.md
AGENTS.md instructions for vercel/next.js, covering next.js development guide, codebase structure, monorepo overview, core package: packages/next and other important packages.
codex AGENTS.md
AGENTS.md instructions for openai/codex, covering rust/codex-rs, the codex-core crate, code review rules, crate api surface and model visible context.
vscode buildNext.instructions.md
Working notes and architecture documentation for the new esbuild-based build system in build/next. Use when making changes to the new build pipeline (transpile/bundle commands, NLS plugin, source-map handling, resource copying, or self-hosting watch tasks).
vscode oss-third-party-notices.instructions.md
Instructions for microsoft/vscode, covering vs code oss third-party-notices pipeline, architecture, pipeline flow in ci, applying the notice (cutover) and fallback chain (never fail the build).
langchain AGENTS.md
AGENTS.md instructions for langchain-ai/langchain, covering global development guidelines for the langchain monorepo, corridor security analysis, project architecture and context, monorepo structure and development tools & commands.
spec-kit AGENTS.md
AGENTS.md instructions for github/spec-kit, covering agents.md, about spec kit and specify, quickstart — add a new integration in 5 steps, integration architecture and integrationmanifest — file tracking.