sc-convex-cloud
Skill13 starsChanged 2 months agoMITView on GitHub
Convex Cloud (managed) deploy operations. Run 'npx convex deploy' against a Cloud deployment using CONVEX_DEPLOY_KEY, output the injected NEXT_PUBLIC_CONVEX_URL, and probe the deployment (/version on *.convex.cloud + JWKS on *.convex.site). The online-path counterpart to /sc-convex (self-hosted on Dokploy).
In this repository28 other files it gives its agents.
- AGENTS.mdAGENTS.md
- si-coderSkill
- sc-allSkill
- sc-axSkill
- sc-buildSkill
- sc-cfSkill
- sc-clerkSkill
- sc-convexSkill
- sc-dokploySkill
- sc-dokuSkill
- sc-dxSkill
- sc-feSkill
- sc-gitSkill
- sc-helpSkill
- sc-installSkill
- sc-n8nSkill
- sc-onboardingSkill
- sc-providerSkill
- sc-resendSkill
- scSkill
- sc-skillSkill
- sc-stripeSkill
- sc-supabaseSkill
- sc-syncSkill
- sc-uiSkill
- sc-uxSkill
- sc-vercelSkill
- use-si-coderSkill
The file
We split this skill into the rules it gives an agent
7 rules. About 0 are common, 1 is distinctive, and 6 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.
6
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 14
Use this skill when the user wants to deploy, debug, or maintain a Convex Cloud (managed) backend — the online-path counterpart to …
- Very specific: line 43
If CONVEX_DEPLOY_KEY is missing, route the user to /sc-onboarding.
- Very specific: line 49
Deploy key is a secret: never log/echo it. Pass it via the process env, never interpolate it into a shell command string.
- Very specific: line 55
Read the main sc delivery workflow for auth/email or domain changes.
- Distinctive: line 58
Do not shell-source dotenv or let ambient self-hosted/dev selectors override the project release credential.
- Very specific: line 72
All scripts live in skills/sc-convex-cloud/scripts/ (repo-relative.
- Very specific: line 75
Run a Convex Cloud deploy.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill 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
/sc-convex-cloud — Convex Cloud (managed)
Use this skill when the user wants to deploy, debug, or maintain a Convex Cloud (managed) backend — the online-path counterpart to /sc-convex (self-hosted on Dokploy). The repo lives at https://github.com/rahmanef63/si-coder-agent.
flowchart LR
A["/sc-convex-cloud"] --> B["scripts/deploy-cloud.js<br/>CONVEX_DEPLOY_KEY (env, never echoed)"]
B --> C["npx convex deploy<br/>→ Convex Cloud"]
C --> D["inject NEXT_PUBLIC_CONVEX_URL<br/>into the build"]
D --> E["probe /version (*.convex.cloud)<br/>+ JWKS (*.convex.site)"]
E --> F["ready ✅"]Connection-safe local execution
For a direct Convex Cloud source=sc named connection, execute credential-dependent helpers through sc run -- ..., for example:
sc run -- node skills/sc-convex-cloud/scripts/deploy-cloud.js --helpThe plain node scripts/... examples below describe the script interface. Do not export CONVEX_DEPLOY_KEY globally merely to satisfy them. Composio-backed connections execute through Composio rather than sc run.
NEVER ask the user to run Convex CLI by hand
Prefer an existing project-owned release wrapper because it owns environment isolation, custom domains and verification. Otherwise use scripts/deploy-cloud.js. Do not instruct the user to run npx convex deploy interactively, nor to hand-set NEXT_PUBLIC_CONVEX_URL. The deploy key is passed via the script's env and never echoed. If a deploy fails, debug it with scripts/check-cloud.js and fix root cause — do not punt the Convex CLI call to the user.
Pre-requisites
CONVEX_DEPLOY_KEY— production (or preview) deploy key for the Cloud deployment. Required.CONVEX_DEPLOYMENT— optional local-dev marker written bynpx convex dev. NOT used in CI.
If CONVEX_DEPLOY_KEY is missing, route the user to /sc-onboarding.
CORE MANDATES (LEARNED LESSONS)
- Respect the project's URL contract: a coupled build may inject a deployment URL, but must not silently replace verified custom API/Auth domains. Read the active frontend adapter and release wrapper first.
- Use the actual adapter's public variable: Next.js uses
NEXT_PUBLIC_CONVEX_URL; other frameworks may usePUBLIC_CONVEX_URLor a compatibility bridge. Set--cmd-url-env-var-nameonly to the actual consumer. - Deploy key is a secret: never log/echo it. Pass it via the process
env, never interpolate it into a shell command string. - Set
CONVEX_DEPLOY_KEYonly on the matching env: a prod key → Production only; a preview key → Preview only. Never setCONVEX_DEPLOYMENTin CI. - Project + key creation is human-one-time: there is no anonymous Cloud-project create API. Mint keys headlessly only after a first interactive login (
npx convex deployment token create <name> --prod --save-env).
Production selection and feature verification
Read the main sc delivery workflow for auth/email or domain changes.
Convex function env is separate from frontend-host env. Verify the exact
deployment and use --prod for production env/run operations that support it.
Do not shell-source dotenv or let ambient self-hosted/dev selectors override the
project release credential. Never dump env values during diagnosis.
Deploy the backend provider code, then verify a production capability query and
the actual UI/provider flow; redeploying only the frontend is insufficient.
Required env vars
| Variable | Format | Purpose |
CONVEX_DEPLOY_KEY | prod:<name>|eyJ2… or preview:… | Authenticates npx convex deploy against the Cloud deployment (SECRET) |
CONVEX_DEPLOYMENT | prod:<name> / deployment name | Optional local-dev marker only; leave blank in CI |
Scripts
All scripts live in skills/sc-convex-cloud/scripts/ (repo-relative; the usage blocks below assume this cwd). On install, this skill is symlinked into ~/.claude/skills/sc-convex-cloud/, so the same paths resolve there too.
deploy-cloud.js
Run a Convex Cloud deploy. Coupled build (default) runs the frontend build via --cmd and injects NEXT_PUBLIC_CONVEX_URL; --backend-only pushes just the backend (when Vercel runs the coupled build itself). Prints NEXT_PUBLIC_CONVEX_URL=<url> — never the deploy key.
node scripts/deploy-cloud.js \
[--build-cmd 'npm run build'] \
[--url-env NEXT_PUBLIC_CONVEX_URL] \
[--backend-only] \
[--message "deploy message"] \
[--cwd <path>]check-cloud.js
Probe a Cloud deployment's /version (on the *.convex.cloud host) + /.well-known/jwks.json (on the *.convex.site host — that's where @convex-dev/auth serves JWKS). Derives the URL from CONVEX_DEPLOY_KEY when --url is omitted. Prints a status table. /version is required; JWKS is advisory (a deployment without @convex-dev/auth has none) and only gates the exit code when you pass --expect-auth. Exits 2 if a required probe is not ok.
node scripts/check-cloud.js --url https://<name>.convex.cloud
# require JWKS too (the deployment is expected to run @convex-dev/auth):
node scripts/check-cloud.js --url https://<name>.convex.cloud --expect-auth
# or, deriving from the prod deploy key:
node scripts/check-cloud.jsDiagnosis
| Symptom | Cause | Fix |
| Client connects to wrong backend | NEXT_PUBLIC_CONVEX_URL not injected | Ensure --cmd-url-env-var-name NEXT_PUBLIC_CONVEX_URL is set on the coupled build |
Browser shows CONVEX_URL unset | Missing NEXT_PUBLIC_ prefix override | Pass --url-env NEXT_PUBLIC_CONVEX_URL (default) — never the bare CONVEX_URL |
JWKS (*.convex.site) not 200 | @convex-dev/auth keys not configured on the deployment (or a probe pointed at the wrong host) | JWKS is served on *.convex.site, NOT *.convex.cloud; configure @convex-dev/auth keys on the deployment via the dashboard / env. Advisory unless check-cloud.js --expect-auth |
deploy-cloud.js exits 1 immediately | CONVEX_DEPLOY_KEY missing | Set it (route to /sc-onboarding) — never run the CLI by hand |
check-cloud.js cannot derive URL | preview key (branch-derived name not in key) | Pass --url https://<name>.convex.cloud explicitly |
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.

