About ask-claude
A skill that runs the local `claude` CLI in non-interactive mode to answer questions and assist with coding. It supports model selection, multiple output formats, and continuing the latest conversation.
Key features
- 01Run queries with `claude -p`
- 02Select a model
- 03Choose text, JSON, or stream-JSON output
- 04Continue the latest conversation
- 05Use the current working directory
Use cases
- 01Ask a coding question
- 02Request a code review
- 03Get structured JSON results
- 04Continue a previous query
Install
Copy the command and run it in a terminal.
$
npx skills add https:/After installing, type/ask-claudein Claude Code to confirm it loaded.
Details
From the project READMEAsk Claude
Executes the local claude CLI to get coding assistance.
Note: This skill requires the claude CLI to be installed and available in your system's PATH.
Quick start
Run a single query with -p (print mode):
claude -p "Your question or task here"
Common options
| Option | Description |
|---|---|
-p | Non-interactive print mode (required for scripting) |
--model MODEL | Specify model (e.g., sonnet, opus, haiku) |
--output-format FORMAT | Output format: text, json, stream-json |
-c, --continue | Continue the most recent conversation |
For all available options, run
claude --help
Examples
Ask a coding question:
claude -p "How do I implement a binary search in Python?"
Use a specific model:
claude -p --model opus "Review this code for potential issues"
Get JSON output:
claude -p --output-format json "List the main functions in this file"
Continue a previous conversation:
claude -p -c "Now add error handling to that function"
Notes
- The
-pflag runs Claude non-interactively and outputs result to stdout - If
--modelis not specified, the default model is used stream-jsonoutputs responses incrementally as they are generated- Warning: The
-pflag skips the workspace trust dialog. Only use in directories you trust. - The command inherits the current working directory
Comments
Markdown supported