Across this series we have met the main command-line flags one at a time. Here they are together, as a quick reference for the ones that quietly earn their keep.
Session and mode flags
-p "prompt"runs headless: print the answer and exit. The door to scripting.-ccontinues the latest session in the project.-r <id> "prompt"resumes a specific session by ID.--permission-mode <default|acceptEdits|plan|bypassPermissions>sets how much Claude asks before acting.
Scope and model flags
--add-dir <path>gives Claude access to folders outside the current project.--modelchooses the model for the run.
claude --model sonnet --add-dir ../shared-lib
Output and prompt flags
--output-format <text|json|stream-json>controls how results print: readable text, structured JSON, or a stream of events.--append-system-prompt "..."adds standing instructions on top of the defaults.
git diff | claude -p "write a commit message" --output-format json
The one to handle with care
--dangerously-skip-permissionsremoves prompts entirely. The name is the warning; keep it to throwaway sandboxes.
Putting it together
The real power comes from combining them. Pipe data in, run headless, ask for JSON, and set a standing rule, all in one line. Once these become muscle memory, Claude Code stops being just an interactive chat and becomes a tool you can wire into everything from a one-liner to a full CI pipeline.
Comments
Be the first to comment.