Ariadne
AGENTS.md4 starsChanged 5 months agoMITView on GitHub
Ariadne is a phone-first event backbone (see README.md). It follows the Dedalus monorepo ethos: minimal diff, deletions celebrated, fail closed, no silent fallbacks, match surrounding style. Process hygiene. pnpm dev, pnpm test, and browser runs are agent-spawned. You own them — tear them down when done. vitest workers here have hung at ~100% CPU for hours across sessions; before ending a turn that started long-runners, kill stray test/watch processes by group (kill -TERM -<PGID>) and keep only the dev server…
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 31,774 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.
The file
We split this AGENTS.md into the rules it gives an agent
2 rules. About 1 is common, 0 are distinctive, and 1 is 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.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Common: line 43
Webhooks fail closed: verify the HMAC signature.
- Very specific: line 46
Read README.md for architecture and skill/SKILL.md for the strap-on contract.
≈ “Common” is an estimate: at least 1 in 5 of the 40,246 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
Instructions for coding agents
Ariadne is a phone-first event backbone (see README.md). It follows the Dedalus monorepo ethos: minimal diff, deletions celebrated, fail closed, no silent fallbacks, match surrounding style.
Commands
pnpm dev # Next dev server on :3939
pnpm typecheck # tsc --noEmit
pnpm test # vitest (run before committing)
pnpm migrate # apply schema to SUPABASE_DB_URL (Supabase Postgres)
pnpm seed # demo data into Supabase via the live brain
pnpm provision # AgentPhone agent + number + webhook
pnpm smoke # live AgentPhone checks
pnpm simulate # signed local inbound webhook
pnpm mcp # strap-on MCP server (stdio)Process hygiene. pnpm dev, pnpm test, and browser runs are agent-spawned. You
own them — tear them down when done. vitest workers here have hung at ~100% CPU for
hours across sessions; before ending a turn that started long-runners, kill stray
test/watch processes by group (kill -TERM -<PGID>) and keep only the dev server
you're actively using. Never kill the IDE or shared MCP servers. Full procedure:
cleanup-terminals-browsers skill / process-hygiene.mdc.
Conventions
- Stack: Next 16 (App Router) · React 19 · TypeScript strict · Tailwind v4 +
cn()· Zod · Vitest · Postgres viapg(Supabase pooler in prod),pglitein tests. Deployed on Vercel. - Structure: constants in
src/constants/, pure logic insrc/domain/, data access insrc/server/db/repositories/(extendBaseRepository), business logic insrc/server/services/, partner adapters insrc/server/partners/. Helpers live in their own files; do not nest functions; imports at top of file. - State is deterministic. Mission pass/fail and drink parsing are decided by code, never an LLM. The agent persona (prompts) only shapes voice.
- Backbone is testable.
new Backbone(await createPgliteDb())wires everything over an in-memory Postgres with no network. Services take deps via the constructor. - CSS: Tailwind utilities only; raw CSS lives in
app/globals.css. Dynamic per-tile gem colors use a single inline-style bridge (documented exception). - Webhooks fail closed: verify the HMAC signature; reject if the secret is unset. Outbound is best-effort and must never block the guest reply path.
Read README.md for architecture and skill/SKILL.md for the strap-on contract.
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.

