菜单
Implement requested code changes end to end from a prompt, spec, or ticket. Use for feature work, diagnosed bug fixes, refactors, and other tasks that require editing code and verifying the result.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Ultra-compressed communication mode. Cuts output tokens 65% (measured) by speaking like caveman while keeping full technical accuracy. Supports intensity levels: lite, full (default), ultra, wenyan-lite, wenyan-full, wenyan-ultra. Use when user says "caveman mode", "talk like caveman", "use caveman", "less tokens", "be brief", or invokes $caveman. Also auto-triggers when token efficiency is requested.
Review a branch, PR, commit range, or working-tree diff along two separate axes: Standards and Spec alignment. Use for review requests and as the quality gate after non-trivial implementation. Load security review only for security-sensitive changes.
Shared vocabulary for designing deep modules. Use when the user wants to design or improve a module's interface, find deepening opportunities, decide where a seam goes, make code more testable or AI-navigable, or when another skill needs the deep-module vocabulary.
Diagnosis loop for hard bugs and performance regressions. Use when the user says "diagnose"/"debug this", or reports something broken/throwing/failing/slow.
Compact the current conversation into a handoff document for another agent to pick up.
Scan a codebase for deepening opportunities, present them as a visual HTML report, then grill through whichever one you pick.
| name | implement |
| description | Implement requested code changes end to end from a prompt, spec, or ticket. Use for feature work, diagnosed bug fixes, refactors, and other tasks that require editing code and verifying the result. |
Implement the work described by the user, spec, or ticket.
Read AGENTS.md and applicable nested instructions first. Inspect the live code and preserve unrelated user changes.
Load karpathy-guidelines and follow it throughout implementation. Use tdd where meaningful, at pre-agreed public seams.
Load only the additional skill that matches the work:
design for UI/UX changes.postgres-patterns for confirmed PostgreSQL work.seo for search-specific work.security-review for trust-boundary or security-sensitive changes.Run typechecking and focused tests regularly, then the relevant broader checks once at the end. Use code-review after non-trivial work.
Do not commit, push, publish, or open a PR unless the user requested it. Load git or handoff only when that action is requested.