Prompt Fixer

The Brief

Reduce friction from rough prompts by cleaning them up before the assistant acts, without ever silently rewriting the user's intent.

The Challenge

Hastily typed prompts, typos, vague one-word asks, filler openers, and broken templates lead to misunderstandings and wasted clarification rounds with AI assistants.

Users should not have to overthink their phrasing, yet assistants respond best to clear, actionable input.

Our Approach

We built a plugin for Claude Code and Antigravity that detects rough prompts and improves them, fixing typos, clarifying vague requests, repairing template syntax, and stripping filler, while always showing the change.

The design is deliberately non-intrusive: it nudges and shows a before and after rather than silently rewriting or blocking anything the user typed.

• On-demand /fix-prompt command with before and after comparison
• Auto-detection hook that flags rough input
• Non-intrusive: never silently rewrites or blocks a prompt
• Customisable detection rules and fixing behaviour

Scope of Work

• ▸Claude Code and Antigravity plugin
• ▸/fix-prompt command with before and after output
• ▸Auto-detection hook (shell + jq)
• ▸Configurable detection and fixing rules

The Outcome

Prompt Fixer catches common prompt issues upfront, giving the assistant clearer, more actionable input and improving response quality without extra effort from the user.

It streamlines any workflow where quick, informal prompts are the norm, cutting the back-and-forth of clarification.

Want results like this? Book a consultation