jev-rules
Claude Code plugin where Jev asks one yes/no per standing rule and delivers only the rules relevant to each prompt, once per session.
Overview
jev-rules addresses the pile of standing instructions that accumulates in Claude Code: test the payment code, use British spelling, follow the deploy checklist. Showing all of them on every prompt makes Claude wade through irrelevant rules, while keyword triggers miss rules when the trigger word is absent. jev-rules asks Jev one yes/no question per rule, whether the request is about that rule, and passes Claude only the rules that get a yes. It takes well under a second, costs a fraction of a cent per prompt, and falls back to showing every rule if anything goes wrong.
Version 0.5.0 adds a line under each prompt naming the rules given and their scores, plus a rules pane beside the conversation that turns rules green with their scores. Earlier versions added once-per-session delivery (so a long session never costs more than loading everything once), rules that follow the file being edited, a filtered codebase map, applies and does-not-apply fields, rule subfolders and retries. It has no dependencies and is MIT-licensed.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Yes / no | Is this request about this rule? | yes · no |
How it works
- On each prompt, send it with all rules to Jev
- Ask one Noul per rule
- Deliver rules that score yes, once per session
- Show chosen rules and scores in the prompt line and pane
- Fall back to all rules on error
Strengths
- Claude sees only relevant instructions
- Fail-safe fallback to all rules
- Transparent per-rule scores
- No dependencies
Caveats
- Requires a TypeSafe key
- Rule wording affects selection quality
Getting started
Install from the repository's Claude Code plugin release and set your TypeSafe key; see the README for setup.At a glance
- Type
- Project
- Status
- beta
- Language
- JavaScript
- Runtime
- Claude Code plugin
- Has code
- yes
- License
- MIT
Use cases and integrations
How curated lists describe it
\*\*New in 0.5.0:\*\* \[one line under each prompt\](#in-the-conversation) names the rules Claude was given and Jev's score for each, and the \[rules pane\](#the-rules-pane) switches on with one answer: the first time Jev picks a rule, Claude asks whether you want it.
logicrw/awesome-jev-projectsScores your standing Claude Code instructions against each prompt and delivers only the ones Jev picks, once per session rather than per message. Ships a pane that shows which rules were chosen.
MrJev/awesome-jev
Frequently asked questions
What does Jev decide?
Whether each standing rule is relevant to the current prompt.
What if Jev fails?
The plugin falls back to showing every rule.
Does it repeat rules every prompt?
No. Since 0.3.0 each rule is delivered once per session.
Related projects
MCP server with record and recall tools: facts are logged during a session and Jev keeps only those relevant to the next task.
MCP serverbetaTypeScriptScores installed Claude Code skills against the current project with Jev and hides irrelevant ones via skillOverrides.
Agent skillbetaJavaScriptPre-message hook that keeps or drops SKILL.md files before the first model call, using local overlap or Jev keep/drop judgments.
HookbetaPythonClaude Code and Codex plugin that removes skills from context and lets Jev decide each turn which SKILL.md bodies to inject.
PluginbetaTypeScriptToolkit where Jev condenses large MCP responses for Claude Code and maps Codex requests to reviewed, fixed repository operations.
ProjectbetaTypeScriptAgent skill that has Jev pick the few relevant lines from big logs before your coding agent reads them, with a reversible reduce and ledger.
ProjectbetaPython