Token导航 LogoToken导航

Agent Skill · 研究检索

implement(实现)

implement 用于查找、检索和筛选相关信息,适合在 Codex、Claude、Cursor、Gemini CLI 中需要根据关键词、任务场景或来源线索快速定位候选结果时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。

implement 是研究检索类 Agent Skill,可用宿主为 Codex、Claude、Cursor、Gemini CLI,通过GitHub 仓库获取,权限为只读;许可证信息还不明确,第三方扫描已通过,按安装量在研究检索类的 73,675 个 Skill 里排第 50,054。

总安装 630周安装 26GitHub Stars 公开资料未说明下载 206核验 2026-05-01

安装说明

本站只整理中文说明,不托管安装包,也不代用户安装。
GitHub
许可证
unknown

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:implement(实现)
来源仓库:https://github.com/kelvinz/cobb
仓库路径:skills/implement
安装命令:
npx skills add https://github.com/kelvinz/cobb --skill implement
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 npx skills 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

skills.shnpx skills
npx skills add https://github.com/kelvinz/cobb --skill implement

这个 Skill 是做什么的

研究检索类 Skill 主要解决「找资料、核对事实、把来源整理清楚」的问题,通常会让助手先检索再总结。

implement 用于查找、检索和筛选相关信息,适合在 Codex、Claude、Cursor、Gemini CLI 中快速定位候选结果。适用于根据关键词、任务场景或来源线索进行信息定位和整理。通过 npx skills add 命令从指定 GitHub 仓库安装使用。

安装前需确认权限范围、维护状态及是否涉及联网或文件操作。implement 属于研究检索类 Skill,可作为该场景下的辅助能力补充。

来源仓库内路径:skills/implement

上手步骤

4 步
  1. 1

    确认自己的宿主工具

    公开资料标注支持 Codex、Claude、Cursor、Gemini CLI;先确认自己用的是其中之一,再按对应工具的 Skills 目录放置。

  2. 2

    拿到安装命令

    可以直接复制的命令是 npx skills add https://github.com/kelvinz/cobb --skill implement;命令里的占位符换成自己的路径或密钥再执行。

  3. 3

    安装后重启宿主

    命令执行完要完全退出宿主工具再打开,让新装的 Skill 被重新扫描到,否则列表里可能看不到。

  4. 4

    按最小权限试用

    它标注为只读,先用只读任务试一次,确认行为符合预期再扩大使用范围。

常见问题

6 条

implement 怎么安装?

在终端执行 npx skills add https://github.com/kelvinz/cobb --skill implement。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

implement 支持哪些 AI 工具?

公开资料标注支持 Codex、Claude、Cursor、Gemini CLI。其他工具只要能读取本地 Skill 目录,一般也能手动放进去,但要自己确认触发方式是否一致。

implement 需要什么权限?

权限级别是只读。该 Skill 主要提供规则、说明或参考内容,本身偏只读;真正读写文件、联网或执行命令仍取决于宿主 Agent 的任务。

implement 是免费的吗?

公开资料没有标注明确许可证,收费和商用限制都无法确认。正式使用前请到来源仓库核对授权说明。

implement 用的人多吗?

公开数据是累计安装 630 次,近一周安装 26 次,下载 206 次。这些数字来自第三方目录统计,只能当作热度参考,不代表质量或安全性。

implement 安全吗?

第三方扫描结果为通过,但这只覆盖已知的静态检查项,不能替代自己读一遍源码和安装命令。

值得选的地方和要先想清楚的地方

值得选的地方

  • 同时适配 4 种工具,换客户端不用重新找替代品。
  • 第三方扫描已通过,常见的高风险行为没有命中。
  • 标注为只读权限,试用阶段不容易误改本地文件。

先想清楚的地方

  • 许可证没有明确标注,能不能商用、能不能再分发都不确定。
  • 第三方目录整理,版本更新和行为细节以来源仓库为准。

适不适合你

更适合

  • 已经在用 Codex,想直接扩展它能力的人。
  • 需要处理研究检索类任务、又不想每次都手写提示词的人。
  • 只想先试试效果、不希望它改动本地文件的人。

不太适合

  • 需要明确授权才能用到公司项目里的场景。
  • 需要厂商级技术支持和 SLA 的生产系统。

研究检索类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
find-skills

寻找技能 · Codex、Claude

31,199,81316,486操作浏览器MIT有提醒
self-improving-agent

自我改进 Agent · OpenClaw

16,464,0023,396接触敏感数据MIT-0通过
soultrace

灵魂追踪 · Codex、Claude

7,588,94412需要联网未明确有提醒
self-improving

自我完善 · OpenClaw

6,829,2841,046需要联网MIT-0通过
microsoft-foundry

微软代工厂 · Codex、Claude

6,614,144764调用外部服务未明确有提醒
implement

实现 · 本页

630—只读未明确通过

按安装量取同分类前几名;数字来自第三方目录统计,只作热度参考,不代表质量结论。

SKILL.md 原文

implement

Implement a feature from a PRD.


Guardrails

  • Follow the repo's AGENTS.md instructions (if present) for any files you touch.
  • Keep changes simple and handoff-friendly: assume a junior dev (or another AI) will maintain this later.
  • Do not change product scope while executing:

- If the PRD is missing details or ambiguous, stop and use prd to refine the PRD first. - If implementation reveals the PRD is incorrect, pause and propose PRD edits via prd. - Do not silently deviate. - If the feature has no PRD in tasks/, stop and use prd first. - If you discover out-of-scope requirements or bugs during execution, do not expand scope. - Create a new PRD via prd instead.

  • Do not reset any existing PRD checklist items when updating an existing PRD.
  • Treat context capture as built-in.
  • When implementation yields durable decisions/gotchas, update tasks/context.md in this step.
  • Use design as an optional companion for UI/UX-heavy work:

- If visual direction, interaction states, or design-token choices are unclear, run design before coding that area. - If approved design artifacts already exist, proceed directly with implementation.

  • Require user confirmation before creating or switching git branches.
  • When requesting user decisions or confirmations, use numbered short-reply options (for example: 1, 2, 3).
  • Do not commit, merge, push, or delete branches; delegate these to commit.
  • Do not claim tests passed or checks succeeded without actually running them; if you didn't run it, say so.

Workflow

  1. Identify the PRD

- Ask for the PRD path if not provided (usually tasks/f-##-<slug>.md). - Confirm which feature ID/name this PRD corresponds to.

  1. Preflight

- Read AGENTS.md (if present) and follow it. - Check priority: - Scan active PRD files in tasks/ for their Status: and Priority: fields. - If higher-priority PRDs with Status: ready exist above this feature (P0 > P1 > P2 > P3), ask the user to confirm working on this item now. - If a higher-priority PRD is still Status: draft, mention it but proceed (it's not ready yet). - Create a new feature branch before making changes (unless you are already on an appropriate feature branch): - Base it off the repository default branch (resolve via origin/HEAD or repo policy). - Propose the branch name and base branch, then ask the user to confirm before creating it. - Name it based on the PRD Summary Type (or match repo conventions): - Type: fix → fix/f-##-<short-slug> - Type: chore → chore/f-##-<short-slug> - default (Type: feat) → feat/f-##-<short-slug> - If tasks/context.md exists, skim key decisions / notes / gotchas relevant to this area before coding. - Parse the PRD into an execution plan. - Include user stories, functional requirements, non-goals, risks, rollout, and testing notes. - Decide whether design is needed: - Trigger it when the PRD adds/changes UI surfaces, interaction/motion behaviour, or design-system patterns. - Also require that no approved design direction is available. - If design runs, treat its output as implementation constraints and keep PRD scope unchanged. - Verify dependencies: - Read this PRD's "Dependencies & Constraints" section for feature dependency IDs. - For each dependency ID, locate PRD files by feature ID in tasks/ and tasks/archive/. - Treat dependencies as complete only when the dependency PRD is archived in tasks/archive/. - If any dependency is still in tasks/, recommend finalising that dependency first. - If a dependency has no PRD yet, stop and use prd to create that dependency PRD first. - Override: if the user confirms dependencies are satisfied, proceed. - Example: dependency work merged but not yet archived. - Record the override in tasks/context.md.

  1. Execute

- Implement the feature as specified. - Implement in small checkpoints aligned to PRD user stories to enable atomic commits. - For UI/UX work, implement against approved design output (states, tokens, layout rules). - Avoid ad-hoc style drift. - Add/update tests as appropriate. - Run the project's normal checks (typecheck/lint/tests/build) per AGENTS.md or project conventions.

  1. Verify

- Verify acceptance criteria and edge cases from the PRD. - Perform any manual QA steps listed in the PRD.

  1. Update checklist progress (in-place)

- Check off completed user stories/tasks and acceptance criteria in the PRD as implementation progresses. - If the PRD lacks checklist items for implementation progress, add a small checklist section and use it. - Do not reset existing checked items.

  1. Update context inline when needed

- Update tasks/context.md when implementation produces durable information: - important gotchas or constraints - architectural/technical decisions with trade-offs - milestone-level completion notes worth preserving

  1. Close out

- Summarise what was changed and what remains. - Next steps: - If unresolved UI/UX direction remains, run design and continue implement. - Run review when needed, then commit in commit mode. - When the feature is ready to merge, run review when needed, then commit in finalise mode.


Output

  • Update code and tests as needed.
  • Update PRD story/task checklist progress in the PRD file.
  • If you discover a durable decision/gotcha, update tasks/context.md in the same run.
  • Also update it when you complete a significant milestone.
  • Reply with:

- PRD path - Which checklist items were completed - Any follow-ups or open issues

  • End with a short status block:

- Files changed: list of created/updated files - Key decisions: any assumptions or choices made (if any) - Next step: recommended next skill or action

平台分布

Codex

35.8%
按下载量换算74

Claude

26.72%
按下载量换算55

Cursor

18.36%
按下载量换算38

Gemini CLI

9.49%
按下载量换算20

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

通过

权限和风险

只读

该 Skill 主要提供规则、说明或参考内容,本身偏只读;真正读写文件、联网或执行命令仍取决于宿主 Agent 的任务。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。目前只有一个公开来源,建议同时核对源仓库。

继续浏览同类 Skills