/verify
Build, launch, and drive the LLM Gateway stack locally to verify a change end-to-end (API + DevPass dashboard + gateway), including Stripe webhook flows.
$ npx -y skills add theopenco/llmgateway --skill verify --agent claude-codeHow it fires
How this skill gets triggered: by you, by Claude, or both.
- Fires itselfAuto-invocation. Claude auto-loads it when your prompt matches the work.Auto-invocation is when the right skill fires by itself at the right moment, driven by a FLOW.md router and a hook, instead of you invoking it by name. It is the difference between a skill being installed and a skill actually getting used.Read the full definition →
- You can call itInvoke it directly when you want it.
- Slash command
/verify
Context preview
The summary Claude sees to decide when to auto-load this skill.
Build, launch, and drive the LLM Gateway stack locally to verify a change end-to-end (API + DevPass dashboard + gateway), including Stripe webhook flows.
SKILL.md
verify.SKILL.mdname: verify
description: Build, launch, and drive the LLM Gateway stack locally to verify a change end-to-end (API + DevPass dashboard + gateway), including Stripe webhook flows.
Verify a change by running the stack
Build first (always via Turbo filters): `pnpm exec turbo run build --filter=api --filter=gateway --filter=code`.
Launch on offset ports (never fight other worktrees for :4002/:3004)
Each server blocks, so run each one in its own terminal session (or as its own background task). In every session, set the shared DB/Redis env first:
# All sessions — persistent containers openllm-pg (5433) and openllm-redis (6380)
export DATABASE_URL="postgres://postgres:pw@localhost:5433/db" REDIS_PORT=6380
Terminal 1 — API on :4102 (env vars on the command line beat ../../.env):
cd apps/api && PORT=4102 API_URL=http://localhost:4102 CODE_URL=http://localhost:3104 \
ORIGIN_URLS="http://localhost:3104,http://localhost:4102" \
node --enable-source-maps --env-file=../../.env dist/serve.js
Terminal 2 — DevPass dashboard on :3104:
cd apps/code && API_URL=http://localhost:4102 pnpm exec next dev --port 3104 --turbopack
Terminal 3 — Gateway on :4101:
cd apps/gateway && PORT=4101 node --enable-source-maps --env-file=../../.env dist/serve.js
If the DB is empty, seed with `pnpm --filter db seed`. Login: `admin@example.com` / password same as email (owns DevPass Pro org `test-personal-org-id`, API key `llmgdev_devpass_test_token`).
Gotchas
- Session auth via curl: POST `/auth/sign-in/email` with `Origin:` header from
ORIGIN_URLS, save cookies (`-c cookies.txt`), reuse with `-b`.
- Dev-plan API keys reject `provider/model` ids on the gateway — use the root
model id (`claude-opus-4-8`, not `anthropic/claude-opus-4-8`).
- Stripe end-to-end: `stripe listen --api-key $STRIPE_SECRET_KEY --print-secret`
(key from root .env; the CLI's own login is expired), restart the API with `STRIPE_WEBHOOK_SECRET_TEST=<whsec>`, then `stripe listen --api-key ... --forward-to localhost:4102/stripe/webhook`. Hosted checkout accepts 4242 4242 4242 4242. Test idempotency with `stripe events resend <evt_id>`.
- Drive the dashboard with Playwright MCP; screenshots must be saved under the
repo's `.playwright-mcp/` (its allowed root).
Read more
name: verify description: Build, launch, and drive the LLM Gateway stack locally to verify a change end-to-end (API + DevPass dashboard + gateway), including Stripe webhook flows.
Verify a change by running the stack
Build first (always via Turbo filters): `pnpm exec turbo run build --filter=api --filter=gateway --filter=code`.
Launch on offset ports (never fight other worktrees for :4002/:3004)
Each server blocks, so run each one in its own terminal session (or as its own background task). In every session, set the shared DB/Redis env first:
# All sessions — persistent containers openllm-pg (5433) and openllm-redis (6380) export DATABASE_URL="postgres://postgres:pw@localhost:5433/db" REDIS_PORT=6380
Terminal 1 — API on :4102 (env vars on the command line beat ../../.env):
cd apps/api && PORT=4102 API_URL=http://localhost:4102 CODE_URL=http://localhost:3104 \ ORIGIN_URLS="http://localhost:3104,http://localhost:4102" \ node --enable-source-maps --env-file=../../.env dist/serve.js
Terminal 2 — DevPass dashboard on :3104:
cd apps/code && API_URL=http://localhost:4102 pnpm exec next dev --port 3104 --turbopack
Terminal 3 — Gateway on :4101:
cd apps/gateway && PORT=4101 node --enable-source-maps --env-file=../../.env dist/serve.js
If the DB is empty, seed with `pnpm --filter db seed`. Login: `admin@example.com` / password same as email (owns DevPass Pro org `test-personal-org-id`, API key `llmgdev_devpass_test_token`).
Gotchas
- Session auth via curl: POST `/auth/sign-in/email` with `Origin:` header from
ORIGIN_URLS, save cookies (`-c cookies.txt`), reuse with `-b`.
- Dev-plan API keys reject `provider/model` ids on the gateway — use the root
model id (`claude-opus-4-8`, not `anthropic/claude-opus-4-8`).
- Stripe end-to-end: `stripe listen --api-key $STRIPE_SECRET_KEY --print-secret`
(key from root .env; the CLI's own login is expired), restart the API with `STRIPE_WEBHOOK_SECRET_TEST=<whsec>`, then `stripe listen --api-key ... --forward-to localhost:4102/stripe/webhook`. Hosted checkout accepts 4242 4242 4242 4242. Test idempotency with `stripe events resend <evt_id>`.
- Drive the dashboard with Playwright MCP; screenshots must be saved under the
repo's `.playwright-mcp/` (its allowed root).
LLM Gateway is an open-source API gateway for Large Language Models (LLMs). It acts as a middleware between your applications and various LLM providers, allowing you to: Route requests to multiple LLM providers (OpenAI, Anthropic, Google Vertex AI, and
Repo: theopenco/llmgateway