steelprompt

steelprompt

Apply Anthropic's official prompt engineering guidelines to any prompt.

Usage

/steelprompt "your prompt to improve"
/steelprompt mode full|preview|ask-only|off

---

Mode switching

If the received prompt matches /steelprompt mode X:

1. Run Bash(echo $CLAUDE_PLUGIN_ROOT) to get the plugin root path, then write {"mode": "X"} to <CLAUDE_PLUGIN_ROOT>/.steelpromptrc using the Write tool with the absolute path
2. Confirm: "Mode changed to X. Active from the next prompt."

Available modes:

Mode	Automatic hook behavior
full	Applies 3-tier framework silently (default)
preview	Same logic as full, shows the engineered prompt before executing
ask-only	Asks only if critical information is missing; does not apply the full framework
off	Hook completely disabled

---

Manual prompt rewriting

If the prompt is vague

(critical context is missing that would significantly change the result)

Use AskUserQuestion with 1–2 concrete questions. Do not proceed further.

Threshold: file unspecified = vague; output type unspecified = vague; "fix" without context = vague.

Chain detection

Before rewriting: does the task explicitly combine two or more sequential operations, connected by "then", "and then", "+", "after", or a comma joining separate actions? Operations: refactor, test, document, deploy, migrate, review, implement, verify, analyze, fix, write, run

Examples that trigger chain detection: "refactor then test", "implement and document", "analyze + fix + deploy" Examples that do NOT trigger: "refactor this function", "test the login flow", "document the API"

If yes — do NOT rewrite as a single prompt. Show a chain plan:

Chain detected (N prompts): -> Prompt 1: [task A — produces: output description] -> Prompt 2: [task B — uses: output of Prompt 1] -> ... (one prompt per operation) -> Prompt N: [final task — uses: all prior outputs]

Then use AskUserQuestion:

• "Run in sequence" — execute Prompt 1 now; user pastes output for next
• "Run as single prompt" — merge into one structured prompt and execute
• "Cancel" — stop

If the prompt is clear or complex

Apply all applicable Anthropic principles:

Principles to apply (all, if relevant):

1. Role — You are a [precise role based on task domain]
2. Context (<context>) — all relevant context BEFORE the task: files, project, environment, known constraints from the conversation. If the task references long files or documents, embed them inside <context> BEFORE the task description — never after.
3. Task (<task>) — imperative mood ("Analyze", "Implement", "Fix"); numbered steps if multi-step
4. Constraints (<constraints>) — what NOT to do, edge cases, limits, style requirements. If the task involves irreversible operations (delete, drop, rm, mv to overwrite, remove, overwrite, truncate, push --force, git reset --hard, destroy, wipe, DROP TABLE, DELETE FROM, ALTER TABLE...DROP), automatically add: (a) explicit confirmation before each destructive action; (b) minimal scope — only touch what is explicitly mentioned; (c) intermediate checkpoints for long operations.
5. Output format (<output_format>) — exact structure: type (code/list/prose), length, sections, format. If the format is rigidly critical (JSON, YAML, SQL), add a prefill note: begin the response with the opening character of that format ({, ---, SELECT).
6. Chain of thought — Think through this step by step before answering for complex or multi-step tasks
7. Examples (<examples>) — 2–3 <example>Input: ... Output: ...</example> if the task is ambiguous or the output format is non-obvious, ONLY if inferable from existing context (conversation, codebase). For ambiguous tasks or where format precision matters, add one <bad_example> alongside each <example> — show what NOT to produce.

Base template:

You are a [precise role].

<context>
[all relevant context: files, project, environment, known constraints]
</context>

<task>
[task in imperative mood, numbered steps if multi-step]
1. ...
2. ...
</task>

<constraints>
[what NOT to do, edge cases, limits]
</constraints>

<output_format>
[exact expected structure: type, length, sections, format]
</output_format>

<examples>
<example>
Input: ...
Output: ...
</example>
</examples>

Think through this step by step before answering.

Omit inapplicable sections. Do not fill with placeholders.

Special cases

• Already XML-structured prompt → refine without dismantling the existing structure
• Atomic prompt → warn: "This prompt is already atomic and self-contained. No engineering needed." — stop here, do not show AskUserQuestion.
• Already well-structured prompt → note what is already good, suggest only real improvements

---

Output format

Everything in the same message, no intermediate text:

1. Show the engineered prompt in a code block, preceded by **Engineered prompt:**
2. One line of explanation ONLY if a choice is non-obvious
3. Use AskUserQuestion with these options:
   • "Run" — execute the improved prompt immediately
   • "Edit" — show raw text without code block, say "Copy, edit, and paste as a new message."
   • "Cancel" — stop