What problem does it solve? Cloud agent platforms hand you a Linux box and a checkout and assume your project's commands just work — an assumption that fails for anything with backing services like Postgres, Redis, CouchDB, or Vault. This Skill reuses your existing devcontainer investment instead of re-deriving environment setup per platform, by splitting the toolchain (a dev image) from backing services (a services-only compose file your boot script runs). ## Core Features & Use Cases - Tier model and fitness assessment: Classify platforms (Ona, Cursor, Copilot, Claude Code cloud, Amp, Devin, Factory, Jules, Cosmos, Codespaces, and more) by adaptation cost and by fitness axes like mission boards and worktree models before committing. - The clouddev.yml contract: A single repo-root manifest declaring prepare/boot/refresh/verify/exec/shell/path/doctor scripts, capabilities, services compose file, and path mappings — with adapters generated from it per platform. - Unattended agent operations: Guidance for running agents in your own CI, including proving the environment before the agent starts, salvaging work from runs that stop early, and fencing paths the agent must not edit. - Use Case: You port a Rails app with Postgres and Redis to Copilot cloud agent: split the compose file, publish a dev image, write prepare/boot scripts, and transcribe the services half into an Actions services: block. ## Quick Start Ask the agent to assess which tier your target cloud agent platform falls into and draft a clouddev.yml manifest with prepare and boot scripts for your devcontainer-based project.