If you're building agents that run in the cloud instead of locally, this is your playbook. It covers the infrastructure patterns that matter: pre-warming sandboxes so users don't wait, snapshot/restore for fast follow-ups, warm pools for high-traffic repos, and the git config gotchas when agents commit code. The real insight here is treating startup latency as the primary enemy, which means doing everything possible at image build time and parallelizing sync with user typing. Also walks through self-spawning agents where one agent can kick off sub-sessions for parallel work, which frontier models can actually handle now if you design the primitives right.
npx -y skills add muratcankoylan/agent-skills-for-context-engineering --skill hosted-agents --agent claude-codeInstalls into .claude/skills of the current project.
Select a file.
metabase/metabase
github/awesome-copilot
UKGovernmentBEIS/inspect_evals
addyosmani/agent-skills