Token导航 LogoToken导航

Agent Skill · 研究检索

ah-finalize-code(啊最终确定代码)

ah-finalize-code 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。

ah-finalize-code 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,权限为执行命令;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 70,176。

总安装 212周安装 9GitHub Stars 公开资料未说明下载 74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:ah-finalize-code(啊最终确定代码)
来源仓库:https://github.com/arinhubcom/arinhub
仓库路径:skills/ah-finalize-code
安装命令:
npx skills add https://github.com/arinhubcom/arinhub --skill ah-finalize-code
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/arinhubcom/arinhub --skill ah-finalize-code

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。

orchestrate 代码最终化流程,包括简化、测试、文档化和审查的全套工作流。适用于发布前的代码质量保障和版本冻结,确保符合规范后再提交 PR。通过 npx 命令安装,默认使用 Opus 模型执行各子步骤并自动提交变更。建议在非主干分支运行,并保留回滚能力以防意外修改。

ah-finalize-code 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/ah-finalize-code

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/arinhubcom/arinhub --skill ah-finalize-code;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

    它需要执行命令,先在测试目录或测试账号里跑,确认没问题再放到正式项目。

常见问题

6 条

ah-finalize-code 怎么安装?

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

ah-finalize-code 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。

ah-finalize-code 需要什么权限?

权限级别是执行命令。安装流程涉及命令执行,可能通过 npx skills add https://github.com/arinhubcom/arinhub --skill ah-finalize-code 联网下载 Skill 或依赖。用户安装前应确认命令来源、仓库内容和执行环境。

ah-finalize-code 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。

ah-finalize-code 用的人多吗?

公开数据是累计安装 212 次,近一周安装 9 次,下载 74 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。

ah-finalize-code 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。

先想清楚的地方

  • 需要执行命令,装到常用环境前要先评估影响面。
  • 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。

不太适合

  • 不能接受 Skill 在你机器上执行命令的人。
  • 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
ah-finalize-code

啊最终确定代码 · 本页

212—执行命令未明确通过

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

Finalize Code

Orchestrate the full pre-PR finalization workflow for the current branch. This involves running a sequence of specialized subagents -- each handling a different aspect of the finalization (simplification, testing, documentation, specs, code review) -- with a commit after every step. The workflow ends by creating a pull request.

Configuration

  • Subagent model defaults: Opus for all subagents except committer, which uses Sonnet
  • Thinking mode: ultrathink effort for all subagents

Procedure

0. Initialize

BRANCH_NAME=$(git branch --show-current)
REPO_NAME=$(basename -s .git "$(git remote get-url origin)")
SAFE_BRANCH_NAME=$(echo "${BRANCH_NAME}" | tr '/' '-')
SPEC_DIR="specs/${BRANCH_NAME}"
PROGRESS_FILE="~/.agents/arinhub/progresses/progress-pr-${REPO_NAME}-${SAFE_BRANCH_NAME}.md"

Read ${SPEC_DIR}/spec.md and extract:

  • BASE_BRANCH from the Base Branch metadata field
  • ISSUE_NUMBER from the Issue Number metadata field

If either value is missing from spec.md, ask the user before proceeding.

Compute the merge base for the current branch:

MERGE_BASE=$(git merge-base "${BASE_BRANCH}" HEAD)

Each subagent that analyzes code changes must compute its own fresh diff before starting its work:

DIFF=$(git diff "${MERGE_BASE}")

This ensures every subagent sees the latest state of changes, including modifications committed by previous steps.

If the repository is a monorepo, identify the correct application from the changed files in the diff and run all commands in the context of that application.

Read the template from references/progress-pr.md, replace all <BRANCH_NAME>, <BASE_BRANCH>, <ISSUE_NUMBER>, and <TIMESTAMP> placeholders with actual values, and write the result to ${PROGRESS_FILE}. Every subagent updates its own section after completing its work.

1. Simplify

Spawn subagent simplifier (Opus, ultrathink):

  • Get fresh diff: git diff "${MERGE_BASE}"
  • Run /simplify with prompt: Simplify changes in current diff only, then check "pnpm preflight"
  • Update ${PROGRESS_FILE} Simplifier section (status: completed, findings)

Then spawn subagent committer (Sonnet) -- run /commit.

2. Retrospective

Spawn subagent retrospective (Opus, ultrathink):

  • Run /speckit.retrospective.analyze with prompt: On the end fix follow-up actions in retrospective file and update retrospective file
  • Update ${PROGRESS_FILE} Retrospective section

Then spawn subagent committer (Sonnet) -- run /commit.

3. Test Creator

Spawn subagent test-creator (Opus, ultrathink):

  • Get fresh diff: git diff "${MERGE_BASE}"
  • Prompt: Run command "npm run test:coverage", find the coverage file location from the npm script test:coverage configuration, read coverage file, add important tests for changes in current diff only, optimize count of tests to avoid redundancy, verify coverage improvements, then run "npm run preflight"
  • Update ${PROGRESS_FILE} Test Creator section

Then spawn subagent committer (Sonnet) -- run /commit.

4. JSDoc Updater

Spawn subagent jsdoc-updater (Opus, ultrathink):

  • Get fresh diff: git diff "${MERGE_BASE}"
  • Prompt: In JSDoc comments on the current diff, add/update spec references using shorthand "SXXX:FR-YYY" or "SXXX:SC-YYY" format (S001=001-minute-qset-react) with short descriptions, remove outdated refs and task numbers (e.g. T012), ensure standard JSDoc format, add missing shorthand to spec.md if needed, and run formatter + linter on modified files.
  • Update ${PROGRESS_FILE} JSDoc Updater section

Then spawn subagent committer (Sonnet) -- run /commit.

5. Tests Docs Updater

Spawn subagent tests-docs-updater (Opus, ultrathink):

  • Get fresh diff: git diff "${MERGE_BASE}"
  • Prompt: Update docs/tests/tests-*.md files to reflect changes in current diff only, ensure that each test file follows the established format and includes all necessary details. If no code changes are detected, do not make changes. Check if there are any redundant or unnecessary tests in changes in current diff -- if so, remove redundant tests.
  • Update ${PROGRESS_FILE} Tests Docs Updater section

Then spawn subagent committer (Sonnet) -- run /commit.

6. API Docs Updater

Spawn subagent api-docs-updater (Opus, ultrathink):

  • Prompt: Run "pnpm docs:generate", then extract type definitions, interfaces, and function signatures from "docs/typedoc" output to create/update "docs/api/index.md" (overview + TOC) and topic files ("docs/api/api-components.md", "api-hooks.md", etc.) with one file per logical section, validate all index links resolve to existing files, and report progress -- do not generate code examples, tutorials, or document private APIs.
  • Update ${PROGRESS_FILE} API Docs Updater section

Then spawn subagent committer (Sonnet) -- run /commit.

7. Spec Optimizer

Spawn subagent spec-optimizer (Opus, ultrathink):

  • Get fresh diff: git diff "${MERGE_BASE}"
  • Prompt: Optimize the "${SPEC_DIR}/" folder by reading all files, removing redundant ones (completed research, moved quickstart guides, done checklists/task lists), consolidating key info into the essential files (spec.md, plan.md, data-model.md, retrospective.md, contracts/), removing obsolete contracts and outdated "Next Steps" sections, and verifying no critical context, valid links, or needed information is lost.
  • Update ${PROGRESS_FILE} Spec Optimizer section

Then spawn subagent committer (Sonnet) -- run /commit.

8. Spec Updater

Spawn subagent spec-updater (Opus, ultrathink):

  • Get fresh diff: git diff "${MERGE_BASE}"
  • Prompt: Update files in the "${SPEC_DIR}/" folder to reflect the code changes in current diff only. Ensure that each spec file follows the established format and includes all necessary details. If no code changes are detected, do not make changes.
  • Update ${PROGRESS_FILE} Spec Updater section

Then spawn subagent committer (Sonnet) -- run /commit.

9. Code Review

Spawn subagent code-reviewer (Opus, ultrathink):

  • Run /ah-review-code with prompt: base branch is ${BASE_BRANCH}, after code review read the code review file and fix all issues you find, then check "pnpm preflight"
  • Update ${PROGRESS_FILE} Code Reviewer section

Then spawn subagent committer (Sonnet) -- run /commit.

10. Create PR

Spawn subagent pr-creator (Opus, ultrathink):

  • Run /ah-create-pr with prompt: base branch: ${BASE_BRANCH}, issue number: ${ISSUE_NUMBER}
  • Update ${PROGRESS_FILE} PR Creator section (status: completed, PR URL).

11. Report to User

Present a summary:

  • Path to ${PROGRESS_FILE} with the full audit trail
  • PR URL
  • List of steps completed and any issues encountered
  • Any unresolved findings from the code review

Important Notes

  • Every subagent except committer runs on Opus with ultrathink effort mode. The committer subagent runs on Sonnet.
  • The ${PROGRESS_FILE} serves as a running audit trail. Each subagent (except committer) updates its section immediately after finishing, so progress is always visible. The committer subagent only creates a commit and does not update the progress file.
  • All Spec Kit output files live in specs/<branch-name>/.
  • The workflow is strictly sequential -- each step must complete before the next begins, because later steps depend on commits from earlier steps.
  • If any subagent fails, note the failure in ${PROGRESS_FILE} and report to the user before continuing. Do not silently skip steps.
  • The /commit command creates a conventional commit with a descriptive message based on the staged changes. The committer subagent should not do anything else beyond creating the commit.
  • Base branch and issue number come from spec.md metadata -- if they are missing, ask the user before starting.
  • Each subagent computes a fresh diff (git diff "${MERGE_BASE}") before starting its work, so it always sees the latest state including commits from previous steps.
  • In a monorepo, identify the correct application from the changed file paths and scope all commands (preflight, test, lint, etc.) to that application.

平台分布

Codex

35.43%
按下载量换算26

Claude

29.1%
按下载量换算22

Cursor

18.88%
按下载量换算14

Gemini CLI

9.16%
按下载量换算7

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

通过

权限和风险

执行命令

安装流程涉及命令执行,可能通过 npx skills add https://github.com/arinhubcom/arinhub --skill ah-finalize-code 联网下载 Skill 或依赖。用户安装前应确认命令来源、仓库内容和执行环境。

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