菜单
Implement a single work-item-spec. Orchestrates five phases (understand, plan, implement, review, finalize), each in a separate sub-agent to keep context clean. Use when the user asks to implement a work item or invokes /implement-feature.
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/jordangaston/claude-plugins --skill implement-feature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。
Use when the user needs multi-source research with citation tracking, evidence persistence, and structured report generation. Triggers on "deep research", "comprehensive analysis", "research report", "compare X vs Y", "analyze trends", or "state of the art". Not for simple lookups, debugging, or questions answerable with 1-2 searches.
Run Charlie Munger's two-track analysis on a decision — first the rational forces that actually govern the outcome, then the psychological biases that may be distorting the judgment. Use this skill whenever the user is making or reviewing a significant decision, asks "am I fooling myself", wants a decision sanity-checked or de-biased, worries they're being manipulated or sold to, or mentions the Munger two-step, two-track analysis, or misjudgment. Especially useful before commitments that are hard to reverse: investments, hires, purchases, partnerships, strategy pivots.
Apply the full Farnam Street latticework of ~100 mental models to analyze a problem or decision. Use this skill whenever the user asks to think through a decision, analyze a problem from multiple angles, stress-test a plan, or explicitly mentions mental models, first principles, inversion, second-order effects, or "how should I think about this". Also trigger for open-ended strategic questions ("should we do X", "what am I missing", "help me reason about this") where structured multi-lens analysis beats a single-frame answer. The skill requires scanning ALL models against the problem, then applying the 3-5 best-fit models deeply.
基于 SOC 职业分类
正在显示 SKILL.md
| name | implement-feature |
| description | Implement a single work-item-spec. Orchestrates five phases (understand, plan, implement, review, finalize), each in a separate sub-agent to keep context clean. Use when the user asks to implement a work item or invokes /implement-feature. |
Implement a single work-item-spec by orchestrating five phases, each in a separate sub-agent with clean context.
Required: A work-item-spec (from /work-item-spec).
/work-item-spec first.Run each phase as a separate sub-agent using the Agent tool. Pass the work-item-spec and prior phase outputs to each agent. Do not start a phase until the previous phase completes successfully.
Spawn a sub-agent with:
phases/understand.md as instructionsExpected output: An understanding summary — what is being built, which files and patterns are relevant, and any risks or ambiguities flagged.
Confirm the understanding summary with the user before proceeding.
Spawn a sub-agent with:
phases/plan.md as instructionsExpected output: A task list created with TodoWrite. Each task maps to one small commit, is independently testable, and is ordered by dependency.
Spawn a sub-agent with:
phases/implement.md as instructionsExpected output: All tasks completed, all tests passing, all changes committed.
Spawn a sub-agent with:
phases/review.md as instructionsExpected output: All code simplified, verified, review issues addressed, changes pushed.
Spawn a sub-agent with:
phases/finalize.md as instructionsExpected output: PR created, user notified with the PR URL.