Slash commands are the control surface for coding agents
The command list matters less than the operating model it enables: plan, constrain, execute, verify, and recover.
Claude Code has many slash commands, but memorizing a catalog is the wrong goal. The better mental model is a development control plane for coding agents: project onboarding, planning, permissions, context management, parallel work, review, verification, remote sessions, and diagnostics.
Group commands by workflow
Basic commands manage state: clear, resume, compact, model selection, project initialization, and permissions. Advanced commands coordinate work: planning, review, background execution, branching, parallel tasks, verification, and debugging.
Practical layers
- Onboarding: initialize project knowledge and working directories.
- Before execution: plan, choose model effort, set permissions.
- During execution: branch, background, parallelize, compress context.
- Before delivery: inspect diff, review, security review, run and verify.
- Recovery: doctor, debug, status, rollback, resume.
The advantage is not that a developer can recite every command. It is that the agent can be operated with discipline: plan, constrain, execute, verify, and recover.
Operating takeaway
The value is not the feature itself. The value is the repeatable operating model it enables when the team has to deliver real work without losing control.