Token导航 LogoToken导航

Agent Skill · 研究检索

roadmap-recommend(路线图推荐)

roadmap-recommend 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。

roadmap-recommend 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 46,860。

总安装 759周安装 31GitHub Stars 公开资料未说明下载 243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:roadmap-recommend(路线图推荐)
来源仓库:https://github.com/kw12121212/auto-spec-driven
仓库路径:skills/roadmap-recommend
安装命令:
npx skills add https://github.com/kw12121212/auto-spec-driven --skill roadmap-recommend
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/kw12121212/auto-spec-driven --skill roadmap-recommend

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。

用于查找和推荐路线图相关资源,支持多方案对比与选择。可根据输入条件筛选出高相关性候选结果,提升决策质量。通过 npx 命令从 auto-spec-driven 仓库安装,操作标准化。使用前应确认是否依赖本地文件或外部 API,避免环境不兼容。

roadmap-recommend 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/roadmap-recommend

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/kw12121212/auto-spec-driven --skill roadmap-re…;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

roadmap-recommend 怎么安装?

在终端执行 npx skills add https://github.com/kw12121212/auto-spec-driven --skill roadmap-recommend。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

roadmap-recommend 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。

roadmap-recommend 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。

roadmap-recommend 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。

roadmap-recommend 用的人多吗?

公开数据是累计安装 759 次,近一周安装 31 次,下载 243 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。

roadmap-recommend 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
roadmap-recommend

路线图推荐 · 本页

759—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

You are helping the user choose the next change from the roadmap and turn the accepted recommendation into a normal spec-driven change.

This Skill's Commands

If you cannot remember the exact command used by this skill, look it up here before running anything. Do not guess.

init: node {{SKILL_DIR}}/scripts/spec-driven.js init
roadmap-status: node {{SKILL_DIR}}/scripts/spec-driven.js roadmap-status
propose: node {{SKILL_DIR}}/scripts/spec-driven.js propose <name>
verify: node {{SKILL_DIR}}/scripts/spec-driven.js verify <name>

Prerequisites

The .spec-driven/ directory must exist at the project root. Before proceeding, verify:

ls .spec-driven/

If this fails, the project is not initialized. Run /spec-driven-init first.

If .spec-driven/roadmap/ is missing, repair the scaffold first:

node {{SKILL_DIR}}/scripts/spec-driven.js init

Steps

  1. Read roadmap context first — before recommending anything, read: You MAY delegate bounded analysis-only work such as candidate comparison, roadmap-context summarization, or likely spec-path discovery to a sub-agent. The parent agent MUST keep the final recommendation, the user confirmation checkpoint, and all proposal artifact writes.

- .spec-driven/config.yaml - .spec-driven/roadmap/INDEX.md - the milestone files relevant to the current phase - .spec-driven/specs/INDEX.md - the relevant main spec files the recommended change is likely to touch - node {{SKILL_DIR}}/scripts/spec-driven.js roadmap-status - the planned change entries for the candidates you are considering as single-line roadmap items in the canonical format - \ - Declared: - - treat milestone declared statuses as limited to proposed, active, blocked, or complete - treat planned change declared statuses as limited to planned or complete

  1. Understand what the user wants optimized — determine whether the user wants the next change chosen for impact, urgency, dependency order, risk reduction, or some other planning goal.
  2. Recommend one candidate change — propose:

- a kebab-case change name - which milestone it comes from - why it is the best next candidate - what alternatives were not chosen yet The recommended candidate MUST already appear under a milestone ## Planned Changes section.

  1. Present a proposal checkpoint — before creating any files, summarize: Then ask for explicit confirmation. If the user wants revisions, continue the recommendation discussion and re-summarize until confirmed.

- the proposed change name - which milestone it comes from - the goal and scope of the recommended change - the main spec areas expected to change - why this item is the best next step - any unresolved questions that would go into questions.md

  1. Resolve open questions before scaffolding — if any unresolved questions remain after the recommendation summary:

- present each open question as a structured block with: - Question - Explanation - Impact - Recommendation - ask the user to answer or confirm the decision needed - Explanation must clarify why the issue is unresolved or what decision is still blocking the proposal - Impact must describe what part of the proposal depends on the answer - Recommendation may suggest a preferred answer, but only as a suggestion - do not treat your recommendation as the resolved answer - do not continue until the user has given an exact answer or explicit confirmation

  1. Scaffold the change after confirmation — run: node {{SKILL_DIR}}/scripts/spec-driven.js propose <name> This creates .spec-driven/changes/<name>/ with the seeded templates.
  2. Fill the five proposal artifacts — after scaffolding, complete the same proposal workflow used by /roadmap-propose: Use the recommended milestone context as planning input, but treat the selected planned change itself as a single-line roadmap item.

- write proposal.md with What, Why, Scope, and Unchanged Behavior - write design.md with Approach, Key Decisions, and Alternatives Considered - populate changes/<name>/specs/ with delta spec files aligned by path with the main .spec-driven/specs/ structure - mirror the main spec path exactly, for example .spec-driven/specs/skills/planning.md becomes .spec-driven/changes/<name>/specs/skills/planning.md - use this canonical sample as the format target: --- mapping: implementation: - path/to/implementation.ts tests: - test/path/to/test.ts --- ## ADDED Requirements ### Requirement: new-capability The system MUST provide <observable behavior>. #### Scenario: success - GIVEN <precondition> - WHEN <action> - THEN <result> ## MODIFIED Requirements ### Requirement: existing-capability Previously: The system MUST <old behavior>. The system MUST <new behavior>. ## REMOVED Requirements ### Requirement: old-capability Reason: This behavior is removed because <reason>. - omit sections that do not apply instead of leaving blank placeholders - if the change has no observable spec impact, leave changes/<name>/specs/ empty rather than creating a prose-only delta file - Do not invent mapping paths when the repository evidence is unclear - include mapping frontmatter in delta spec files when implementation and test paths are knowable from repository context - write tasks.md using this canonical structure: ` # Tasks: <change-name> ## Implementation - [] Describe the first atomic implementation task - [] Describe the second atomic implementation task ## Testing - [] Run npm run lint — lint or validation task - [] Run npm test — unit test task ## Verification - [] Verify implementation matches proposal scope Each task uses - [] checkboxes and should be independently completable. The ## Testing section MUST satisfy these verification keyword requirements: If the relevant command cannot be determined confidently from repository context, record that in questions.md instead of guessing. - At least one task MUST contain a lint/validation keyword: lint, validate, validation, typecheck, type-check, or build - At least one task MUST contain a unit test keyword: unit test or unit tests - Both tasks MUST name an explicit runnable command in backticks or use a known runner (npm, pnpm, yarn, bun, node, bash, sh, pytest, jest, vitest, go, cargo, make, uv, poetry) - Paraphrasing these keywords (e.g. "run tests" instead of "run unit tests") will cause verify to fail - write questions.md, recording every unresolved point under ## Open, or leave <!-- No open questions -->` if nothing is unclear

  1. Validate before presenting the proposal — run: node {{SKILL_DIR}}/scripts/spec-driven.js verify <name>

- Fix any safe artifact-format issues immediately and rerun verify - If only open questions remain, treat that as expected at proposal time and surface those questions clearly - If any non-question error remains, stop and report it instead of presenting the proposal as ready

  1. Offer the execution handoff — show the user the generated artifacts, summarize the final proposed scope, list any open questions that must be answered before implementation, and ask whether they want to:

- enter /spec-driven-apply <name> for the stepwise execution path - enter /spec-driven-auto for the end-to-end execution path Do not auto-enter either execution path without the user's explicit choice.

Rules

  • Do not implement code — this skill is planning only
  • Use roadmap context and roadmap-status output as the basis for recommendation
  • Explain why the recommended change should come next
  • Recommend only work that already exists under Planned Changes
  • Do not scaffold proposal artifacts until the user explicitly confirms the recommendation summary and change name
  • If open questions remain, ask the user to resolve them before scaffolding
  • For each open question, provide Question, Explanation, Impact, and Recommendation
  • If testing commands are not knowable from repository context, record that as an open question instead of inventing commands
  • Recommended answers do not count as question resolution without explicit user confirmation
  • After confirmation, follow the same artifact-writing and validation standard as /roadmap-propose
  • Do not let a sub-agent scaffold the change or own the recommendation checkpoint
  • Keep implementation and test mappings in spec frontmatter, not in requirement prose

平台分布

Codex

35.62%
按下载量换算87

Claude

29.34%
按下载量换算71

Cursor

18.16%
按下载量换算44

Gemini CLI

9.57%
按下载量换算23

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