Load only the rules and skills that apply to each agent prompt
Ask Jev which of your team rules, skills or guideline sections apply to the current prompt and inject only those into the agent's context.
The use case
Teams that give coding agents long rule files and many custom skills find the agent ignores rules buried in a wall of text and wastes tokens loading everything. These plugins ask Jev, for each prompt, which rules, skills or guideline sections apply, and inject only those. One loads just the relevant parts of Apple's design guidelines. A developer keeps the full rule set in the repository, reviews which items get picked over a week of work, and adds always-on rules for anything safety related so they are never filtered out.
Example workflow
- Split rules and guidelines into short named sections
- On each prompt, ask Jev a Choice or Noul question per section
- Inject only the selected sections into Claude Code's context
- Mark safety rules as always on
- Review selection logs weekly
Tools
- Jev (TypeSafe AI)
- Claude Code
- Pi
Risks and guardrails
- A needed rule left out leads to a silent mistake
- Section names that are vague get mis-selected
- Extra calls add latency to each prompt
Sources
Written by Agents Lookup. Credits and official pages for the tools this use case names:
- jev-rules by Elia Alberti on Made with Jev: madewithjev.com/builds/jev-rules
- jev-rules: original post or repo by Elia Alberti: github.com/EliaAlberti/jev-rules
- Skill suggestion injects one skill by Shweta Bajaj on Made with Jev: madewithjev.com/builds/skill-suggestion-injects-one-skill
- Skill suggestion injects one skill: original post or repo by Shweta Bajaj: x.com/shwetabjaj/status/2101917086469759060
- AskHIG by NatashaTheRobot on Made with Jev: madewithjev.com/builds/askhig
- AskHIG: original post or repo by NatashaTheRobot: x.com/NatashaTheRobot/status/2106569281651834950