菜单
Capture important architectural decisions, bug patterns, feature designs, or other critical findings into the project knowledge base. Use proactively when significant technical decisions are made, non-obvious bugs are resolved, new subsystems are designed, or important product features are discussed.
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
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。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/astashov/liftosaur --skill kb命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
Build a feature end to end — decide worktree vs here, design the architecture and debate it with codex, get the user's approval on a one-pass-readable plan, state the API/state contract, implement with visible diffs, then run two codex reviews (bugs + architecture) and document the result. Use when asked to build/implement/add a feature or do a refactor that involves design decisions.
Create, run, debug and tear down an isolated dev worktree — its own branch, node_modules, domains, ports, SSL certs, DNS records, Metro and iOS simulator. Use when asked to set up/remove a worktree, when something in a worktree points at the wrong port/host/simulator, or when work should happen off the base checkout.
Generate or refresh an architecture walkthrough doc (lambda/scripts/archdocs/*.md) for a subsystem or feature branch — top-down text walkthrough with verified clickable file:line links, mechanically extracted public API, invariants, and a debugging map. Use when asked to map/document/explain the architecture of a subsystem, feature, or branch, or to refresh an existing archdoc after changes.
基于 SOC 职业分类
正在显示 SKILL.md
| name | kb |
| description | Capture important architectural decisions, bug patterns, feature designs, or other critical findings into the project knowledge base. Use proactively when significant technical decisions are made, non-obvious bugs are resolved, new subsystems are designed, or important product features are discussed. |
| disable-model-invocation | false |
| argument-hint | ["what to capture"] |
Capture this into the knowledge base: $ARGUMENTS
All files live in lambda/scripts/memory/. The index is at lambda/scripts/memory/INDEX.md.
Read lambda/scripts/memory/INDEX.md to understand what's already captured and avoid duplicates.
| Category | Directory | When to use |
|---|---|---|
| Architecture | architecture/ | System design, infrastructure, sync protocols, data formats, tech stack choices |
| Decisions | decisions/ | "We chose X over Y because Z" — explicit tradeoff analysis |
| Features | features/ | Product feature descriptions, UX designs, user-facing behavior |
| Bugs | bugs/ | Non-obvious bug root causes, debugging patterns worth remembering |
| Future Plans | futureplans/ | Planned work not yet started — roadmap items, RFCs for future features |
| SEO | seo/ | SEO research, keyword analysis, backlink data, Core Web Vitals |
If none fit, create a new category directory and add a section to INDEX.md.
Filename: lowercase-kebab-case, descriptive. Example: watch-echo-loop-bug.md
Entry format:
---
date: YYYY-MM-DD
tags: [relevant, tags]
---
# Title
Concise description of the finding/decision/pattern.
Keep it 5-20 lines. Link to relevant source files where useful.
Focus on the WHY and the non-obvious parts — assume the reader is a developer familiar with the codebase.
Add a one-liner entry under the appropriate category heading:
- [filename.md](category/filename.md) — concise one-liner (max 80 chars)
Keep entries sorted by relevance within each category (most important first), not by date.
Capture knowledge when ANY of these happen during a conversation:
Do NOT capture: