marketing-skills / app
AGENTS.md73 starsChanged 3 days agoMITView on GitHub
This project deploys to Graphed cloud from graphed.yaml. It is UGC Studio: a Next.js app. The page is React. The JSON API and the render worker run in that same server. Kysely talks to Postgres in the cloud and to embedded PGlite when DATABASE_URL is unset, so npm test needs no database server. - Never guess manifest fields or CLI flags. Run graphed docs manifest or graphed <command> --help first. - Every vendor call is a Graphed Tool (packages/core/src/tools.ts →…
- Reads credentials
- Installs packages
Where files are headingNew AGENTS.md files have more rules about keeping secrets out of the code, and fewer about code style.
Of the AGENTS.md files we read, the 32,671 first seen in the last six months, against the older ones: the kind of rule that gained most share, and the kind that lost most. About files like this in general, not this one, and an estimate, since a kind is a pattern in the words.
In this repository14 other files it gives its agents.
- skills/ad-spy/app/AGENTS.mdAGENTS.md
- skills/enrichment-tables/app/AGENTS.mdAGENTS.md
- skills/post-engagers/app/AGENTS.mdAGENTS.md
- ad-spySkill
- analytics-dashboardSkill
- enrich-linkedin-engagersSkill
- enrichment-tablesSkill
- feedbackSkill
- post-engagersSkill
- post-to-videoSkill
- static-adsSkill
- tesseract-motionSkill
- ugc-studioSkill
- ugc-videoSkill
The file
We split this AGENTS.md into the rules it gives an agent
3 rules. About 1 is common, 0 are distinctive, and 2 are very specific to this project.
≈1
Common · estimate
Many files like this have a rule of this kind.
≈0
Distinctive · estimate
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
- could be automatic: a setting could enforce it
- Very specific: line 9
Never guess manifest fields or CLI flags: Run graphed docs manifest or graphed <command> --help first.
- Common: line 11
Never add vendor API keys.
- Very specific: line 16
Do not read
process.envin app code: Add the key to the zod schema in packages/core/src/config.ts.
≈ “Common” is an estimate: at least 1 in 5 of the 41,581 AGENTS.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
AGENTS.md
This project deploys to Graphed cloud from graphed.yaml. It is UGC Studio: a Next.js app. The page is React.
The JSON API and the render worker run in that same server. Kysely talks to Postgres in the cloud and to embedded
PGlite when DATABASE_URL is unset, so npm test needs no database server.
Golden rules
- Never guess manifest fields or CLI flags. Run
graphed docs manifestorgraphed <command> --helpfirst. - Every vendor call is a Graphed Tool (
packages/core/src/tools.ts→@graphed-inc/sdk) or the Graphed LLM proxy. Never add vendor API keys. - Rendering spends credits.
POST /api/videosandnode cli.mjs renderrefuse to run withoutconfirm: true/--yes. Only send that after the user approved the estimate. - Schema changes are Kysely migrations in
packages/core/src/migrations/, registered indb/migrate.ts.ensureSchemais idempotent. Never edit a migration that has already been applied; add a new file. - Do not read
process.envin app code. Add the key to the zod schema inpackages/core/src/config.ts. - One database.
initDb()/getDb()inpackages/core/src/db/index.ts. - The worker runs inside the Next server (
packages/dashboard/boot.ts). It is not a cron job. - Media never streams through the app. The API returns presigned api.graphed.com links. Presigned links can live 1 hour at most.
- Grid geometry is set from React (a ref writes
transform/ size). Do not put column or row geometry in astyle=attribute.app/studio.cssis the page layout.
Layout
packages/dashboard/Next.js App Router.app/page.tsxrenders the studio.app/api/[...path]/route.tsis the JSON API.packages/core/src/database, worker, render pipeline, plans, research, writer.cli.mjsterminal access (graphed exec --project <slug> -- node cli.mjs …). It reads and queues. The web worker renders.engine/python word fixer (needs ffmpeg + python3, both in the Dockerfile).seed/avatars/<slug>/built-in avatars (avatar.jpg+avatar.json),seed/voice-ref.mp3the shared voice sample.
Checks
npm test (offline, fake vendors), node uitest.mjs (real Chrome, needs npm i --no-save puppeteer-core),
graphed exec --project <slug> -- node smoke.mjs (cloud, no spend).
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
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.

