AgentSkillsCN

railway-ops-autofix

通过 CLI 和 Web 控制台（浏览器 MCP）为 GradeOS 项目运行 Railway：监控日志、检测部署/构建/运行时错误、更新环境变量、清除构建缓存，并重新部署。当用户希望查看 Railway 日志、修复 Railway 部署失败，或为 GradeOS 执行 Railway 控制台操作时，可选用此技能。

SKILL.md
--- frontmatter
name: railway-ops-autofix
description: "Operate Railway for the GradeOS project via CLI and web console (browser MCP): monitor logs, detect deploy/build/runtime errors, update environment variables, clear build cache, and redeploy. Use when the user asks to watch Railway logs, fix Railway deployment failures, or perform Railway console operations for GradeOS."

Railway Ops Autofix

Overview

Monitor the GradeOS Railway project via CLI and the web console, detect errors, and apply fast fixes that are explicitly allowed (env var changes and cache clears), then verify recovery.

Workflow Decision Tree

0) Scope & safety

  • •Target project is GradeOS. Default environment is production unless the user states otherwise.
  • •If the environment or service is unclear, ask once; otherwise pick the most recently failing service in logs.
  • •Allowed without confirmation: update env vars, clear build cache, redeploy.
  • •Anything destructive or structural (delete services/envs/volumes, scale to zero, change plans, remove domains) requires explicit user confirmation.

1) Connect & open project

  • •CLI: confirm context with railway status. If unlinked, run railway link and select GradeOS.
  • •If needed, link a specific service or environment via railway service and railway environment.
  • •Web console: open the project dashboard (use railway open or direct navigation) and select GradeOS.
  • •If any login or 2FA prompt appears, stop and ask the user to complete it.

2) Monitor logs (CLI + web)

  • •CLI: use railway logs to view the latest deployment logs. Use -d for deployment logs and -b for build logs; --json if you need machine parsing.
  • •Web console:
    • •Build/Deploy panel for a specific deployment's logs.
    • •Observability > Log Explorer for environment-wide logs and filters.
  • •Keep the web log stream open for real-time monitoring; use the CLI for quick snapshots after each fix.

3) Detect errors

  • •Look for crash loops, build failures, missing env vars, or connection errors.
  • •Optional classifier: pipe log text into scripts/scan_logs.py for a fast pattern match, e.g. railway logs | python scripts/scan_logs.py.
  • •If the error is unclear, capture:
    • •Service + environment
    • •Deployment/build ID
    • •Timestamp and 20-50 lines around the error
    • •Recent config changes (if any)

4) Fix actions (auto)

  • •Missing env var: set it via CLI (railway variables --set "KEY=VALUE", add -s/-e if needed) or via the Variables UI, then redeploy if required.
  • •Clear build cache: set NO_CACHE=1 for the service, redeploy, then remove NO_CACHE after a successful build.
  • •Redeploy: use railway redeploy -s <SERVICE> or the Deployments UI to re-run the latest deployment.
  • •Keep fixes minimal: only env vars or cache changes are allowed without user confirmation.

5) Verify & iterate

  • •Re-check logs immediately after redeploy and again after the service has been up for a few minutes.
  • •If the same error persists, gather fresh context and apply the next likely fix.
  • •If a fix would require destructive changes or plan changes, ask the user first.

Resources

scripts/

  • •scripts/scan_logs.py: lightweight log classifier (stdin or file input) to highlight common error types.

references/

  • •references/railway-ops.md: CLI + web console actions for logs, variables, and redeploys.
  • •references/error-patterns.md: common log signatures mapped to the safest next action.