Could Jev

A Claude skill that answers one question fast: could Jev help with what I am building? Ask something like 'could jev help here?' or 'is this a jev thing?' and you get a short, structured design gut-check rather than an implementation. It is MIT-licensed, by rlfordon.

It does not call Jev. It reasons about whether Jev fits the problem, using a factsheet and a pattern catalog, and returns a sketch of what the state, questions and code would look like. The README calls it a design gut-check.

In Claude Code, run /plugin marketplace add rlfordon/could-jev then /plugin install could-jev@could-jev, or copy skills/could-jev/ to ~/.claude/skills/could-jev/. For Claude.ai or Desktop, download could-jev.zip from the latest release and upload it under Settings, Capabilities, Skills.

About ten lines: a verdict, the patterns that might apply with their tags, a sketch of state, questions and code, the closest prior art, what to watch out for, the cheapest test to try and a bigger idea if it works.

The obvious classify-X use plus funnels in front of an LLM, censuses instead of samples, locators, paragraph heat maps, pairwise judgments, uncertainty spotlights and agent tripwires, and experimental image checks and Box localization since Jev 1.14.

A monthly sweep checks TypeSafe's docs and releases and looks for new things people are building, then opens a PR with a dated report in maintenance/sweeps/. Update the plugin with /plugin marketplace update could-jev to pick up a refresh, and the skill will say when the factsheet's checked date looks stale.

Logging needs filesystem access, so it works best in Claude Code; saying 'log this jev result' appends experiments to ~/.claude/could-jev/history.md, outside the skill folder, so updating the plugin never overwrites your history or standing rules. On Claude.ai or Desktop the skill still answers, but the log is unavailable.