JevPromptCoach
Claude Code plugin that logs your prompts locally and later has Jev score how well you prompt the coding agent over time.
Overview
JevPromptCoach scores how well you write prompts to a coding agent and shows whether your habits are improving. Unlike tools that score each prompt before it is sent, it adds no latency: its UserPromptSubmit hook appends one line to a local file and exits. Scoring runs only when you invoke a command, using Jev.
It can also score prompts you have already written: the README estimates that a backfill over a year of local Claude Code history costs about six cents, so you get a report on day one. The plugin bundles its dependencies (376 KB), contacts only api.typesafe.ai, and stores the TypeSafe key in a locked-down file under ~/.claude rather than the shell profile. It requires Node 22+ and Claude Code 2.1.x or newer. Docs are available in English, French and Spanish. It is an unofficial community plugin, not affiliated with TypeSafe or Anthropic.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Score | How well written is this prompt to the coding agent?(inferred) | — |
| Yes / no | Does this prompt show a particular good or bad prompting habit?(inferred) | — |
How it works
- Hook logs each submitted prompt to a local file
- User runs the scoring command when desired
- Jev scores logged prompts against the coaching rubric
- Report shows scores and trends over time; backfill can score past history
Strengths
- Zero added latency on each prompt
- Cheap backfill over existing history
- Only contacts api.typesafe.ai; no runtime fetches
Caveats
- Unofficial; requires your own TypeSafe key
- Rubric details are not described in the README excerpt
- Depends on Claude Code hook behaviour in 2.1.x+
Getting started
claude plugin marketplace add CrowdLinker/JevPromptCoach && claude plugin install jevpromptcoach@jevpromptcoach, then put TYPESAFE_API_KEY in ~/.claude/jevpromptcoach/.env.At a glance
- Type
- Plugin
- Status
- beta
- Language
- TypeScript
- Runtime
- Claude Code plugin (Node 22+)
- Has code
- yes
- Has tests
- yes
Use cases and integrations
How curated lists describe it
Claude Code plugin that scores how well you prompt a coding agent, and shows whether your habits are improving. Runs on TypeSafe's Jev model. Zero added latency.
hellogumbo/awesome-jev
Frequently asked questions
Does it slow down my prompts?
No. The hook only appends a line to a local file; scoring happens later on demand.
Can it score past prompts?
Yes. The README estimates about six cents to backfill a year of Claude Code history.
Is it official?
No, it is an unofficial community plugin.
Related projects
sort where the key is a description: Jev judges pairs of texts and a Bradley-Terry fit turns them into a scored ranking.
ProjectbetaPythonChrome extension that grades readable page sections on ten rubric scores with Jev via Vercel AI Gateway and gives the page an A-E grade.
ProjectbetaTypeScriptCLI and agent skill that checks every line of a document against every rule with Jev and outputs flags for an AI agent to fix.
Agent skillbetaTypeScriptTournament engine that ranks texts with Swiss pairing, Elo and TrueSkill, using Jev or local models to judge each head-to-head duel.
ProjectprototypePythonWindows tray proof of concept that re-picks Bopomofo homophone candidates for a whole sentence with Jev before you press Enter.
ProjectprototypeC#CLI that transcribes a video, has Jev score every sentence against rubric flags, and renders a 16:9 edit with a live meter.
AppbetaPython