jiaweixie/jev-enhanced-plugin
Third-party skills enhanced with advisory TypeSafe Jev judgments: two-axis code review, prose humanizing (en/zh-TW), reply checking, simplification gating, and design-tree grilling, plus a procedure for migrating further skills.
Jev-checked variant of code-review. Two-axis review (repo Standards + Spec) of a working tree, PR, branch, or diff since a commit, tag, or merge-base, where TypeSafe Jev checks each finding's quoted evidence and its basis on its own axis, and never drops a finding. Use when the user asks to review code, a PR, a branch, or uncommitted changes, asks "what changed since X", asks whether an implementation matches its spec or issue, or says 幫我 review 或 code review; prefer it over an unaugmented code-review skill. Not for fixing bugs or writing code.
Jev-checked variant of code-simplifier. Behavior-preserving cleanup of recently modified or named code (cut redundancy, flatten nesting, clarify names), verified by tests plus a TypeSafe Jev risk check for dropped cases, changed failure paths, and contract drift. Use when the user asks to simplify, clean up, tidy, de-duplicate, reduce nesting, or refactor code for readability, or to polish code written earlier in the session (簡化、整理、重構程式碼); not for new features, bug fixes, or performance work.
Jev-checked variant of grilling. Stress-test a plan, design, or decision through a dependency-tree interview that asks only questions whose prerequisites are settled, recommends an answer to each, and waits per round; TypeSafe Jev flags questions the conversation already answered. Use when the user says "grill me", "poke holes", "challenge my plan", "stress-test this", "what am I missing", or 幫我挑毛病、質疑我的計畫; not for writing or implementing the plan.
Jev-checked variant of humanizer for English. Rewrite AI-sounding prose so every claim survives while not-X-but-Y contrasts, one-line closers, triads, dash overuse, hype, stock AI words, and bold labels go; TypeSafe Jev flags passages that say nothing, repeat, overclaim, or an editor would still reword. Use when asked to humanize, de-AI, or make English text sound less like AI, or when polishing English emails, docs, articles, READMEs, issues, PRs, or comments; for Traditional Chinese use s1-humanizer-zh-tw.
Jev-checked variant of humanizer-zh-tw. Rewrite AI-sounding or translationese Taiwan Traditional Chinese (zh-TW) prose, cutting 進行/針對/透過 filler, mainland terms, triads, hype, and dash overuse while keeping every claim; TypeSafe Jev flags passages a Taiwanese editor would still reword. Use when asked to 潤稿、改寫、去 AI 味、修翻譯腔、讓文字更自然, or to edit zh-TW articles, docs, announcements, or release notes; for chat replies use s1-i-have-adhd-zh-tw, and not for watermark or invisible-Unicode removal.
Jev-checked variant of i-have-adhd-zh-tw. Reply in natural Taiwan Traditional Chinese (zh-TW) that opens with the answer, the result, or the one action the user must take, uses Taiwanese terms, keeps code, paths, and APIs verbatim, and drops openers, recaps, and sign-offs; an optional TypeSafe Jev check reviews long replies for answer-first, delegation, and coverage. Use when replying to the user in zh-TW for coding, debugging, planning, research, or chat (繁體中文回覆、台灣用語、先講結論); to edit an existing zh-TW article use s1-humanizer-zh-tw.
Port an existing agent skill into the jev-enhanced-plugin repository as an s1- skill, with a license audit, a judgment pack of TypeSafe Jev questions (Noul, Choice, Score) with its model-state builder, the adapted SKILL.md, live wording checks, tests, and docs. Use when the user asks to migrate, port, or adapt a skill into this plugin, or to add or improve a Jev/System One judgment pack here; for TypeSafe in other applications use the typesafe-ai skill, and not for unrelated new skills.