jev-skill-suggestion

Name: jev-skill-suggestion
Author: claude-code-templates

jev-skill-suggestion · by claude-code-templates

Picks one skill per prompt and loads it while hiding the listing from context; uses TypeSafe's Jev with a key, else Claude's classifier.

❯ /plugin marketplace add SeongGwangJu/k-mods
❯ /plugin install jev-skill-suggestion@k-mods

$ claude plugin marketplace add SeongGwangJu/k-mods
$ claude plugin install jev-skill-suggestion@k-mods

Run /reload-plugins to apply it to an open session. Needs Claude Code 2.1.287 or later.

Notes

If you enter an API key (typesafeApiKey or gatewayApiKey), the full prompt, the candidate skills' names and descriptions, and the first part of the selected skill's SKILL.md are sent to the TypeSafe server (a fact the author states directly in code comments and the README).
Without a key, it calls Claude Code's own model classifier ($.model.classify) on every prompt and uses your Claude usage.
Every failure path fails open. If a request fails or times out (default 800ms), it passes the prompt through with no suggestion.
/jev-skill-suggestion:setup does not write the config file itself. It hands Claude a change plan as a prompt, and Claude edits through its Edit tool with your confirmation.

Mods run unsandboxed with your permissions. These labels come from static analysis.

NetworkCalls a modelReads conversationReads env/settingsReads files

Static analysis

Events handled

prompt.attachment{type=skill_listing}prompt.submitsession.compactsession.endskill.promptskill.prompt{skill=jev-skill-suggestion:setup}

Environment variables read

HOME