菜单
Manage Architecture Decision Records and ADR-aligned development. Use when user says "we should use", "let's go with", asks "why did we choose X?", implements a feature that may have an ADR, updates ADR status, or needs to check alignment with existing decisions.
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
Query code semantics (symbols, hover, definitions, references, diagnostics) across any language via the lsmcp MCP server, which bridges through kakehashi to real language servers (lua_ls, basedpyright, rust_analyzer, r_language_server, nixd, …). Use when exploring or analyzing a codebase, finding/understanding symbols, or when the first project-wide search misses a language. Prefer these tools over Grep/Read for code analysis.
Run the full multi-stage review pipeline on the current branch — /review (multi-perspective subagents) → codex MCP → PR → Copilot/Gemini bots — fixing findings at each stage until convergence. Use when the user asks to "revise", run the review pipeline, or prepare a branch/PR for merge review.
Run a strict multi-perspective review with fact-checked findings. Use when the user asks for a review, code review, design review, risk review, or wants independent reviewer perspectives before merging or shipping.
Perform deep multi-perspective analysis with problem-specific analytical lenses, cross-validation, and a final recommendation. Use when the user asks to ultrathink, deeply analyze a complex problem, compare strategic options, expose risks, or make a high-impact technical decision.
Stage meaningful diffs and create Conventional Commits with WHY-focused messages. Use whenever making git commits, grouping changes, choosing commit types, or separating structural and behavioral work.
Guide structural code improvements using Kent Beck's Tidy First methodology. Use when seeing messy code, before making behavioral changes, after completing features, documenting deferred cleanup, or discussing when to clean up code.
| name | adr |
| description | Manage Architecture Decision Records and ADR-aligned development. Use when user says "we should use", "let's go with", asks "why did we choose X?", implements a feature that may have an ADR, updates ADR status, or needs to check alignment with existing decisions. |
Manage Architecture Decision Records (ADRs) throughout their lifecycle: read existing decisions, write new ones, and ensure development aligns with documented decisions.
template.md for the MADR structure (status, context, drivers, options, consequences)operations.md for directory discovery, naming conventions, commit integration, deprecation (_ prefix), and ADR-aligned development workflowqueries.md for finding ADRs by topic, status, or relationshipTrigger: "why did we choose X?", "what's our approach to Y?", "is there an ADR for Z?"
operations.md)Trigger: "we should use", "let's go with", "I've decided to", trade-off discussions
template.mdTrigger: Implementing features, making architectural changes, deviating from existing patterns
_ prefixBefore finalizing an ADR:
Based on MADR - Markdown Architectural Decision Records (Version 4.0, 2024)