home/productivity/prompt-fix

Prompt Fix

GPTClaudeDeepSeek··572 copies·updated 2026-07-14
prompt-fix.prompt
---
description: Upgrade a specific flagged prompt to follow best practices. Proposes a diff and applies it only after you approve.
argument-hint: "<file:line of the prompt, or a description>"
allowed-tools: Read, Glob, Grep, Edit, Agent
---

You are fixing the prompt at / described by: **$ARGUMENTS**

1. If `$ARGUMENTS` is empty or ambiguous, ask the user which prompt to fix (a
   file:line from a prior `/prompt-audit` is ideal).

2. **Re-review the target** via the `prompt-reviewer` subagent to get the current
   findings and a proposed rewrite (so the fix is grounded in the rubric, not ad hoc).

3. **Check placement.** Using the `prompt-best-practices` architecture rules, determine
   whether the prompt is in the right layer/location. If not (e.g. it sits in a domain
   entity or a controller), include a move/extraction in your proposal.

4. **Present the diff** in the terminal with a short rationale tied to rubric ids.

5. **Apply only on approval.** After the user approves, use `Edit` to apply the change.
   If the fix involves extracting/relocating the prompt, create the new file and update
   the call site. Then summarize what changed.

Never edit before the user approves the diff.

when to use it

Community prompt sourced from the open-source GitHub repo assist-software/claude-code-repository (NOASSERTION). A "Prompt Fix" style prompt — adapt the placeholders and specifics to your task. Imported as-is and not independently retested here, so check the output before relying on it.

tags

productivitycommunitydeveloper

source

assist-software/claude-code-repository · NOASSERTION