How do I install the Deployment skill?

Run npx skillmds@latest add cleanexpo/deployment in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

What does the Deployment skill do?

Deployment constraints and runtime behaviour for Railway (backend) and Vercel (frontend) — proxy config, redirects, Telegram, SSE limits, and model routing. It is listed under Web & Frontend on SkillMD.

Is Deployment safe to use?

This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Deployment?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Deployment free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Deployment?

cleanexpo (@cleanexpo) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Deployment by cleanexpo — Agent Skill

Name: Deployment
Author: cleanexpo

File contents

Deployment Patterns

When to apply

Apply whenever touching Railway config, Vercel SSE routes, HTTP fetchers, Telegram notifications, or per-phase model selection.

Rules

Railway terminates TLS and proxies all requests. Do NOT add TrustedHostMiddleware restricting to 127.0.0.1 — it blocks all cloud traffic. Use ALLOWED_ORIGINS for CORS control instead.
Anthropic docs URLs (docs.claude.com) redirect to platform.claude.com and code.claude.com. Any httpx fetcher that hits Anthropic docs must set follow_redirects=True or requests will return 3xx and no content.
Telegram push from any sandboxed Python environment needs only TELEGRAM_BOT_TOKEN + a chat_id. The full python-telegram-bot package is NOT required — urllib + a POST to api.telegram.org/bot{token}/sendMessage is sufficient. See scripts/send_telegram.py.
Vercel hard-limits SSE responses to 300 seconds. The analysis pipeline must abort in-flight SDK calls at 240 s and send a partial done event before the platform cuts the connection. Use AbortController — a budget timer alone cannot stop a running SDK call.
Use per-phase model routing to stay within the 300 s Vercel SSE window: haiku-3-5 for inventory and summarisation phases (1, 2, 4) and sonnet for intelligence phases (3, 5, 6, 7). This cuts total analysis runtime from ~350 s to ~200 s.

Frequently asked questions

Deployment