jev-scoutlisted

Inspect the current repository and find where TypeSafe Jev decision models would create the most value - replacing repeated LLM judgments and brittle heuristics, adding checks where code takes the first match, a fixed cutoff or a default, and new product capabilities that become practical when typed judgments are fast and cheap. Use when asked to "Jevify" a codebase, find Jev use cases or opportunities, reduce AI cost or latency, add typed routing, scoring or verification, or explore what a system could do with frequent inexpensive decisions. Produces a value map, a decision inventory, ranked opportunities in three families (substitution, augmentation, new capability), Jev Choice/Noul/Score decision contracts and shadow-mode experiments. Analysis only - it changes no code and writes nothing outside its report unless the user asks.
PedroAlvarado/jev-scout · ★ 0 · AI & Automation · score 75

Install: claude install-skill PedroAlvarado/jev-scout

# Jev Scout Find where Jev can create material product or business value in the repository the coding agent is working in. Work from evidence in the repository, not from a generic list of Jev use cases. The skill is host-neutral: it works the same in Claude Code, Codex and other Agent Skills hosts, in any ordinary repository. Look for three families of opportunity, and give the third the same effort as the first: 1. **Substitution:** an existing model call or keyword/regex rule makes a bounded judgment that Jev can make faster, cheaper or more reliably. 2. **Augmentation:** a decision is made implicitly today - the first match wins, a fixed top-k cuts the list, everyone gets the same default, an action is not checked, a check runs only at the end, a person reviews everything - and an explicit, typed judgment would make it better. 3. **New capability:** a product or operational behavior the system does not attempt today because a judgment per item, per step or per user was too slow or too expensive, and that becomes practical with fast, cheap typed decisions. Substitutions are the easiest to see and usually the smallest. The biggest value is often in the other two families. ## Boundaries - **Analysis only.** Do not edit code. Write nothing outside the report: no issues, tickets, comments, commits, pull requests, or records in any tracker or coordination tool - even when the repository's agent instructions (`AGENTS.md`, `CLAUDE.md` or similar) tell agents to file what the