Prompt Sensei v1.0.0

[chengzhongwei]

Prompt Sensei v1.0.0

A local-first prompt coach for AI coding agents (Claude Code and Codex). It teaches better prompting habits one at a time, without telemetry or raw prompt storage.

What it does

• Observe — adds a one-line score after each of your prompts (stage, score, and a concrete next-habit tip), so you see calibration in real time without leaving your editor.
• Improve — rewrites a single prompt into a more executable version while preserving your intent.
• Lookback — with separate consent, analyzes your recent local Claude Code or Codex history and turns it into a coaching report.
• Report — a private, local summary of your scores, repeated gaps, and the habit to practice next.
• Setup / settings — guided consent, auto-start scope, and optional redacted previews via structured pickers in the host.

How it runs

• Works in Claude Code (slash commands) and Codex (natural language).
• Optional host-native auto-start hooks on both: Claude Code uses SessionStart / UserPromptSubmit / Stop / PreCompact; Codex uses the same three minus PreCompact. Hooks install at folder or user scope and stay inert until you opt in.

Privacy

Local-first by default: no cloud, no telemetry, no raw prompt text stored. Only timestamps, scores, stages, and tip metadata; redacted prompt previews are opt-in.

Install

git clone https://github.com/chengzhongwei/Prompt-sensei ~/.claude/skills/prompt-sensei
(cd ~/.claude/skills/prompt-sensei && npm install && npm run build)

Then /prompt-sensei observe in Claude Code, or Use prompt-sensei observe mode. in Codex.

---

This discussion was created from the release Prompt Sensei v1.0.0.