Claude Prompt

---

Type /claude-prompt:p <your request> and Claude reads it, rewrites it into an optimized prompt — adding structure, specificity, and meta-instructions that get more out of extended thinking — then carries it out. You see the rewrite, then you get the work. Clear asks run in one shot; only genuinely ambiguous asks get a quick clarifying question first, and high-stakes (destructive/irreversible) asks get a one-line confirm before anything runs.

It applies four techniques:

1. Structured context — explicit reasoning frameworks and step-by-step structure.
2. Specificity — vague asks become detailed requirements with clear success criteria.
3. Meta-instructions — guidance that leverages extended thinking and planning.
4. Skip-comments — literals are reproduced verbatim: "double quotes", `inline code`, fenced code blocks, file paths, URLs, regexes, and error strings survive the rewrite untouched.

It also types intent (do a task · answer a question · improve a piece of text) and supports a --dry optimize-only mode. No code. No dependencies. No network. No hooks. Just one auditable command. MIT.

Install

In Claude Code, add the marketplace and install the plugin:

/plugin marketplace add danielkremen818/claude-prompt
/plugin install claude-prompt@claude-prompt

That's it — the command is now available as /claude-prompt:p. Prefer the shorter /p? See Shorten to /p below. To update later, the bump happens automatically on the next /plugin sync once a new version is published.

Requires Claude Code with plugin support. See the official plugin docs if /plugin isn't recognized.

Usage

After install, the command is /claude-prompt:p. Run it followed by whatever you'd normally ask:

/claude-prompt:p add input validation to the signup form

Prefer the shorter /p? Set up the one-line alias in Shorten to /p.

Claude responds in two parts — the rewritten prompt, then the work:

Example: quoted text is never touched

Anything inside double quotes is preserved exactly — labels, copy, identifiers, error strings:

/claude-prompt:p rename the CTA to "Get started — it's free" everywhere

The optimizer restructures the task but leaves "Get started — it's free" byte-for-byte:

Just the prompt — --dry

Prefix --dry to get only the optimized prompt, without executing it — handy when you want to copy the rewrite somewhere else or review it first:

/claude-prompt:p --dry migrate the test suite from mocha to vitest

When to use it

• Vague or one-line asks you want Claude to scope properly before acting.
• Tasks where you want explicit success criteria and a plan, not a guess.
• Anything where a sharper prompt would mean a sharper result.

For trivial, well-specified requests, skip it — the rewrite step is overhead you don't need.

Shorten to /p