القائمة
Review a diff for method violations: rot to rewrite, translate-only layers, unvouched deps, fancy over brute force. One line per finding.
يُحدَّد Prompt الذي يراجع المصدر أولًا بشكل افتراضي. يمكنك التبديل إلى أمر مباشر أو تنزيل نسخة محلية.
اقرأ SKILL.md وأي ملفات مرافقة يعرضها SkillsMP قبل أن تقرر التثبيت.
التثبيت باستخدام Codex أو Claude انسخ هذا Prompt والصقه في Codex أو Claude أو مساعد آخر ليراجع صفحة Skill ويثبّتها لك.
يتجاوز الأمر المباشر Prompt المخصّص للمراجعة. افحص المصدر قبل تشغيله.
npx skills add https://github.com/rajnandan1/ken --skill ken-reviewيبقى الأمر في سطر واحد. مرّر أفقيًا لمراجعته كاملًا قبل النسخ.
تفضّل نسخة محلية؟ نزّل الملفات المتاحة حاليًا لدى SkillsMP.
Thompson-mode discipline for any coding task: think first, build bottom-up, brute force until measured, rewrite over patch. Not for non-coding requests.
Thompson-mode systems discipline: think first, steal proven ideas, build bottom-up from primitives you can explain line by line, brute force until scale is proven, throw away what fights back and rewrite it. Supports intensity levels: lite, full (default), ultra. Use on ANY coding task: writing, adding, refactoring, fixing, reviewing, or designing code, and choosing libraries or dependencies. Also use whenever the user says "ken", "ken mode", "thompson mode", "brute force it", "rewrite it", or "throw it out and redo it". Do NOT use for non-coding requests (general knowledge, prose, translation, summaries, recipes).
Audit the whole repo for Thompson-mode violations. A ranked list of what to rewrite, delete, or take back into the trusted base.
عرض SKILL.md
| name | ken-review |
| description | Review a diff for method violations: rot to rewrite, translate-only layers, unvouched deps, fancy over brute force. One line per finding. |
| homepage | https://github.com/rajnandan1/ken |
| license | MIT |
Review diffs for Thompson-mode violations. One line per finding: location, what's wrong, the move. The diff's best outcome is fewer layers and a smaller trusted base.
L<line>: <tag> <what>. <the move>., or <file>:L<line>: ... for
multi-file diffs.
Tags:
rot: unit on repeated patches; the next fix is a rewrite, not patch N+1. Name the rewrite's shape.layer: wrapper, adapter, or manager that adds no decision of its own. Delete it, move callers down a level.unvouched: dependency or pasted code nobody present can explain. Vouch for it (read it) or replace with a few lines you own.fancy: clever algorithm where the plain loop wins at this scale. Name the brute-force form; demand the measurement that justifies fancy.ceremony: process, abstraction, or config that serves the process itself. Remove it.❌ "This caching layer might benefit from reconsidering its abstraction boundaries and perhaps simplifying the invalidation strategy."
✅ L12-60: rot: third patch on this cache's invalidation. Rewrite on a plain dict + mtime check, ~25 lines.
✅ L4: unvouched: left-pad-like microdep imported unread. 3 lines inline, trusted base shrinks by one.
✅ svc.py:L88: layer: OrderManager delegates every call to OrderRepo. Delete, callers hit OrderRepo.
✅ L30-52: fancy: hand-rolled B-tree for 40 entries. Flat array + linear scan; ken: revisit when n > 10k measured.
✅ L71: ceremony: config flag nobody sets gates one constant. Inline the constant.
End with the metrics that matter: net: -<N> lines, -<M> layers, trusted base -<K> deps possible.
If there is nothing to flag, say Sound. Ship it. and stop.
Scope: Thompson-mode method violations only. Correctness bugs, security
holes, and performance are out of scope. Route them to a normal
review pass, not this one. A single smoke test or assert-based self-check
is the ken minimum, not ceremony, never flag it for removal.
Does not apply the fixes, only lists them.
"stop ken-review" or "normal mode": revert to verbose review style.