craft-prompt

name	craft-prompt
description	Craft copy-paste-ready prompts. Use to write prompts, turn notes into templates, draft `/goal` conditions, or answer "프롬프트 만들어" requests.

craft-prompt

Purpose

Turn a goal, scattered notes, or a raw ask into a clear, well-structured prompt that gets the best results from any LLM — a separate step from doing the task itself, so the result is a copy-paste-ready text block the user can drop into any AI interface.

Use this when

• the user asks to write, make, or build a prompt from scratch
• scattered notes need to become a reusable prompt or template
• a session handoff prompt or /goal condition is needed
• Korean prompt requests appear, such as "프롬프트 만들어" or "프롬프트 작성"

Inputs

• Goal — what the prompt should make the LLM do
• Target — which model, product surface, or agent interface
• Audience — who will use the prompt (the user, a team, non-technical users)
• Reuse — one-shot use, or a reusable template with placeholders
• For coding-agent or worktree prompts: the execution context — default to paths relative to the current worktree root unless the user explicitly needs machine-specific absolute paths

Don't over-ask — infer whatever the request already makes clear. "Write me a code review prompt for GPT" already gives goal and target.

Workflow

1. Confirm inputs. Ask only what Inputs doesn't already make clear.

2. Gather the raw material. Collect what the prompt's purpose needs — task instructions and constraints; research questions and recency needs; for a session handoff, run git status, git diff --stat, git log --oneline -5, and note recent changes; for a /goal condition, outcome/evidence/constraints/non-goals/scope/budget/stop condition; for a system prompt, persona/capabilities/boundaries/tone; for a template, variable placeholders and usage notes. Distill rough notes or a conversation dump rather than repeating them verbatim.

3. Build the prompt, using only the blocks it needs:

   Block	When to include
   Role	A specific expertise improves the output (one sentence)
   Context	The LLM needs background to reason correctly
   Task	Always — the core instruction
   Rules / Constraints	There are important do's/don'ts
   Output format	The default output structure won't work (see Output format)
   Examples	Desired behavior is hard to describe but easy to show

   Sizing heuristic: a small request (under ~20 words of user specification) stops at Role + Task + Rules; add Output format only if the user names a shape, Examples only if behavior is hard to describe. A reusable template turns every value that varies week-to-week into a {{placeholder}} — hardcoding more than one varying value under-templates.

   For research, coding-agent, or other high-impact prompts, add a small verification contract — what to check before finalizing: requirements, grounding, format, irreversible side effects — instead of generic "be careful" prose.

   XML tags work across major LLMs; use them for complex or multi-section prompts, plain text/markdown for simple ones. If the user names a target model, product, image generator, or video generator, load the matching guide or template (below) and keep volatile target-specific behavior out of this portable spine.

4. Sharpen. Cut fluff (no sentence that fails to change LLM behavior); be specific (numbers beat adjectives); keep it self-contained (no dangling "this conversation" references); resolve hidden conflicts between rules, examples, and edge cases; state how missing context should be handled (look up, ask, or proceed with labeled assumptions). Weigh every cut against Principles — especially Cut in this order and Right-sized. For complex prompts, run through references/quality-checklist.md for deeper failure-mode analysis.

5. Deliver. See Output format.

Output format

A fenced code block, ready to copy-paste — always, when the user explicitly asked for a prompt or clearly invoked this skill. Never replace the prompt with direct task execution.

If relevant, add a brief note outside the code block: how to customize placeholders, which parts to adjust if results aren't ideal, and recommended target-platform settings (model choice, search focus mode). Ask if the user wants adjustments; refine tone, blocks, or target LLM based on feedback.

Non-English prompts: write the prompt body in the requested language, but keep XML tag names in English (<context>, <task>) — all major LLMs parse English tags regardless of content language.

Guardrails

• always deliver the prompt when one was asked for, even if the underlying task looks simple enough to do directly
• keep target-specific formatting quirks in guides/ and templates/, not in this spine
• default to worktree-relative paths in coding/worktree prompts; state the base once if it could be ambiguous
• don't inflate a simple prompt to look thorough — see Principles

Principles

Shared with craft-tune — full statements in references/shared-principles.md:

• Context beats instruction
• Outcome over process
• Cut in this order
• Right-sized beats thorough-looking

Prompt-specific:

1. The prompt is the product. Deliver polished text the user copies as-is, not a meta-discussion about prompting.
2. Respect the target LLM. XML tags work broadly, but each interface has nuances — read the target's guide when available, especially for search-first, multimodal, or local-model workflows.
3. Reusability when asked. Use {{placeholder}} syntax with clear labels for templates; bake in specifics for one-shot prompts.
4. Verification beats vague caution. For complex or high-impact prompts, say what to verify before finalizing instead of piling on broad "be thorough" instructions.
5. Know when to skip. Skip prompt-building only when the user didn't ask for a prompt and direct execution is clearly better; if they explicitly asked for a prompt, deliver it even when the underlying task is simple.

Failure modes

• process worship — listing every step to reach the outcome instead of stating the outcome and letting the target LLM find its own path
• template worship — templating a one-shot request nobody will reuse, or the reverse: hardcoding a value that actually varies week to week
• missing output contract — delivering prose about the prompt instead of a copy-pasteable fenced block, or dropping a shape constraint the user named
• fluff inflation — padding with generic "be helpful/thorough" language that doesn't change behavior
• format mismatch — ignoring the named target's conventions, or leaking platform-specific instructions into this portable spine instead of guides/

Example

Input: "write me a code review prompt for GPT, keep it short"

Delivered prompt:

You are a senior code reviewer.

Review the diff below for correctness bugs, security issues, and unnecessary complexity. Skip style nits.

Report only issues you're confident about, one per line: `file:line — issue — suggested fix`.

<diff>
{{diff}}
</diff>

Note outside the block: "Swap {{diff}} for the actual diff before sending."

References (load on demand)

• references/shared-principles.md — full statements of the four principles shared with craft-tune
• references/components-guide.md — Deep dive on each building block with examples and anti-patterns
• references/prompt-patterns.md — 8 common patterns: research, code gen, review, writing, extraction, analysis, handoff, decision
• references/quality-checklist.md — Quality checks with failure modes and fixes
• references/goal-conditions.md — Writing /goal completion conditions and reviewable goal specs for Claude Code and Codex autonomous loops (transcript-visible evidence, cross-platform differences, and caveats)

Guides (platform-specific tips)

• guides/claude-guide.md — Claude Code / Claude web / API
• guides/gpt.md — GPT Pro / ChatGPT / API
• guides/perplexity.md — Perplexity
• guides/gemini.md — Gemini / Imagen
• guides/local-models.md — Local/open-source models (Llama, Qwen, Mistral, DeepSeek, Ollama, etc.)

Templates (special cases where a well-crafted template adds real value)

• templates/session-handoff.md — AI session handoff (continuation, debug)
• templates/image-gen.md — Image generation (photo, illustration, icon, per-platform notes)
• templates/video-gen.md — Video generation (text-to-video, image-to-video, camera keywords, per-platform notes)
• templates/system-prompt.md — Chatbot/agent system prompts (layered architecture)