Setup

On first use, read setup.md for integration guidelines. If local memory is needed, ask for consent before creating ~/render-deploy/.

When to Use

Use this skill when the user wants to deploy, publish, or host an application on Render and needs reliable deployment execution instead of generic advice. Activate for render.yaml Blueprint generation, MCP direct service creation, runtime configuration checks, and post-deploy triage.

Architecture

Memory lives in ~/render-deploy/. See memory-template.md for setup.

~/render-deploy/
|- memory.md                  # Stable preferences and integration choices
|- deployment-notes.md        # Project-level deployment decisions
|- env-inventory.md           # Required env vars and source of truth
- incident-log.md            # Deploy failures and resolved fixes

Quick Reference

Load only the minimum file needed for the current request.

• RENDER_API_KEY is exported in the shell, or
• Render CLI is authenticated (render whoami -o json)

For git-backed flows, require git and a valid remote URL. Do not attempt opaque credential discovery or unrelated environment inspection.

Core Rules

1. Classify the Deployment Path First

• git remote -v for source availability
• MCP availability (list_services())
• CLI fallback readiness (render --version, render whoami -o json)
• Active workspace context (MCP or CLI)
• Authentication presence (RENDER_API_KEY or authenticated CLI session)

Do not proceed to deployment steps when prerequisites are missing.

4. Treat render.yaml as Executable Infrastructure

• Declare all required env vars
• Mark user-provided secrets with sync: false
• Prefer plan: free unless user requests another plan
• Match service type and runtime to the actual app behavior

After creating the file, validate before push.

5. Require Push Before Deeplink Handoff

• Confirm latest deploy status is live
• Check health endpoint response
• Review recent error logs
• Validate required env vars and port binding (0.0.0.0:$PORT)

If failures exist, run one-fix-at-a-time triage from troubleshooting.md.

Common Traps

• Starting deploy without a git remote -> Blueprint and MCP git-backed flows fail immediately.
• Picking Direct Creation for multi-service systems -> Missing workers/datastores and fragmented setup.
• Forgetting sync: false on secrets -> Broken deploys or accidental secret exposure in config.
• Using localhost binding instead of 0.0.0.0:$PORT -> Health checks fail even when process is running.
• Redeploying repeatedly without root-cause fix -> Noisy failures and delayed resolution.

External Endpoints

• Preferences and deployment history in ~/render-deploy/ if the user accepts memory.
• Local codebase inspection outputs and interim analysis notes.

This skill does NOT:

• Read unrelated credentials outside the deployment context.
• Scrape credentials from shell history, dotfiles, or unrelated config paths.
• Send project files to undeclared third-party endpoints.
• Run destructive infrastructure changes without explicit confirmation.

Trust

By using this skill, deployment metadata and selected configuration are sent to Render services. Only use it if you trust Render with this operational data.

Related Skills