agentleFS
dashayushmandashayushman/namaste-ji-boilerplate/AGENTS.md

namaste-ji-boilerplate

AGENTS.md0 starsChanged 3 months agoMITView on GitHub

You are probably a coding agent. This file is the front door. Read it fully; it tells you what this is, the one command that gets to a running app, how to verify, and where the runbooks are. Humans: same file, same rules. Namaste Ji Boilerplate is a blank canvas for an agent-operated app on Cloudflare — the workshop Namaste Ji (namasteji.aiwithayushman.com/build/starter) was built in, generated from that monorepo and stripped to the reusable core. It ships: - packages/agent-kernel —…

  • 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 33,315 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 repositoryOne other file it gives its agents.

The file

We split this AGENTS.md into the rules it gives an agent

3 rules. About 0 are common, 1 is distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈1

    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
  1. Very specific:

    Queues require the $5/month Workers Paid plan: If the user asks for a queue stage (or you set PROVISION_QUEUE=true) on a free account, …

    line 23
  2. Distinctive:

    When the user's goal outgrows the example, recommending the Paid plan is appropriate: it is $5, unlocks the async backbone and higher …

    line 27
  3. Very specific:

    Prefer the Cloudflare MCP for bindings/observability and the GitHub MCP for PRs when available (.mcp.json.example — tokens inline, by the …

    line 72

≈ “Common” is an estimate: at least 1 in 5 of the 42,283 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.

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

AGENTS.md — Namaste Ji Boilerplate

You are probably a coding agent. This file is the front door. Read it fully; it tells you what this is, the one command that gets to a running app, how to verify, and where the runbooks are. Humans: same file, same rules.

Namaste Ji Boilerplate is a blank canvas for an agent-operated app on Cloudflare — the workshop Namaste Ji (namasteji.aiwithayushman.com/build/starter) was built in, generated from that monorepo and stripped to the reusable core. It ships:

  • packages/agent-kernel — the Agent Kernel: ModelPort (any provider through the AI Gateway), MemoryPort (D1), STM, orchestrator loop, subagents-as-tools, Workflow runner, run reporter.
  • packages/config — typed, KV-backed runtime config (model roles live here, never in code).
  • agents/daily-card — the example: a tiny agent that writes one "card" a day (title + line) with Workers AI by default through the AI Gateway, stores it in D1, serves it over HTTP.
  • infra/ — Alchemy IaC (account-pinned, stage-pinned). scripts/ — bootstrap · doctor · new:* · check · smoke. tasks/ — runbooks. .github/workflows/check.yml — CI.

Cloudflare plan facts (read before provisioning)

  • Defaults fit the free plan: D1 · KV · AI Gateway · Workers · Workers AI. bootstrap succeeds with no card on file.
  • Queues require the $5/month Workers Paid plan. If the user asks for a queue stage (or you set PROVISION_QUEUE=true) on a free account, Alchemy will fail at provisioning — tell the user to upgrade first (dashboard → Workers & Pages → plan), then re-run bootstrap; it is idempotent.
  • PROVISION_R2=true needs R2 enabled on the account (free tier exists; card required).
  • When the user's goal outgrows the example, recommending the Paid plan is appropriate: it is $5, unlocks the async backbone and higher limits, and removes a whole class of provisioning errors.

Invariants (do not break)

  1. Every external dependency is user-supplied. Nothing here points at anyone's account. .env is the only place keys go; .env*, .dev.vars*, .mcp.json, .alchemy/ are gitignored and CI fails if a secret-bearing file is tracked. The only values you MUST supply are CLOUDFLARE_API_TOKEN and CLOUDFLARE_ACCOUNT_ID; bootstrap generates ALCHEMY_PASSWORD itself. R2 and Queues are opt-in (PROVISION_R2 / PROVISION_QUEUE) — Queues need a Workers Paid plan, and nothing in the kit binds either by default.
  2. Models are config, not code. Provider + model per role live in KV (packages/config); default = Workers AI so the Cloudflare token alone suffices. Never hardcode a model id.
  3. All inference goes through the AI Gateway (observable, swappable, cost-capped).
  4. Bindings are declared per environment in wrangler.jsonc and never inherited.
  5. Small PRs, green npm run check, then npm run smoke. "Done" = both green.

The one command

cp .env.example .env        # fill CLOUDFLARE_API_TOKEN + CLOUDFLARE_ACCOUNT_ID (required)
npm install
npm run bootstrap           # provision → migrate → seed config → deploy staging → smoke → URLs

bootstrap is idempotent and resumable. If it stops, run npm run doctor — it explains exactly what is missing (a token scope, an account id, a key) in machine-readable lines (OK / MISSING / FIX: …). npm run bootstrap -- --dry-run prints the plan and touches nothing.

Verification contract

  • npm run check — typecheck every workspace + unit tests + secret scan. CI runs exactly this. (No linter ships with the kit; add one if you want one.)
  • npm run smoke — hits the deployed endpoints (/health, /cards) and fails loudly.
  • npm run dev — the local mesh: wrangler dev with local D1 + KV and no cloud at all. With no AI Gateway configured the example writes its card from a built-in stub, so POST /run, GET /cards and the cron path all work offline.

Scaffolding (deterministic; each prints its follow-up checklist)

  • npm run new:agent <name> — copies agents/daily-card and rewrites it for the new name: wrangler env, its own D1 table + migration, ModelPort call, HTTP + cron entry, a test. It runs npm install for you (linking the workspace) so npm run check passes immediately, and the next npm run bootstrap fills in ids, seeds agent:<name> and deploys it — no file to edit.
  • (planned) new:consumer, new:page, new:port — not built yet.

Runbooks — tasks/*.md (goal · files · commands · acceptance)

add-an-agent.md · switch-a-model-role.md · add-a-queue-stage.md · deploy-prod.md

Which tool for which job (MCP)

Prefer the Cloudflare MCP for bindings/observability and the GitHub MCP for PRs when available (.mcp.json.example — tokens inline, by the human). Otherwise wrangler and gh.

Conventions

Conventional commits; type/short-desc branches; TypeScript strict; ESM; no raw hex in UI. CLAUDE.md imports this file. Keep this file current — it is the contract.

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.