Deploy [service name] to [cloud environment] and then check the live logs rather than relying only on the tests. Look at the CloudWatch or equivalent output and confirm that our custom logger is actually producing…
Add an end-to-end latency benchmark for [MCP server] before changing any runtime behavior. Measure cold and warm calls for health checks, tool listing, tool execution, and search, and compare…
Review [batch pipeline path] and find service startup, shutdown and coordination code that a single bounded job could replace. Replace it with a job runner that runs the media processing step, such as…
Convert this Jenkinsfile into a GitHub Actions workflow for [repository name]: [paste Jenkinsfile]. Keep the same stages, environment variables and artifacts, and open a pull request with…
Help me troubleshoot [device problem] on a remote [device type]. I have an SSH session open in a tmux pane named [pane name]. Use that pane to run read-only diagnostics first, such as checking logs,…
Read the goal-file template and examples at [goal file template link] to understand how a GOAL.md works. Then write a GOAL.md for this repository that defines a weighted composite score for [test health goal, such…
Plan [feature name] with me first, listing the scope, risks and files likely to change. Once I approve, act only as coordinator: break the work into tasks, assign each one to a sub-agent, and run…
In [demo environment], inspect the cache metadata exposed by [MCP server]. Propose small configuration changes, such as TTL adjustments, and apply them only within [demo approval policy]. Record each suggestion…
Work through the ready items in [task queue] with one Claude Code lane per task. Give each lane its own Git worktree and implement the change there. When a lane finishes, have a separate review…