scan-new-specs

作者: warpdotdev

扫描warpdotdev/warp和warp-server中最近合并的PRODUCT.md规范,这些规范尚未在warpdotdev/docs中有对应的文档PR。当发现完整规范时,自动生成完整的文档草稿PR并标记工程师。当规范过于简略无法起草时,直接通知工程师。设计为定时运行的Oz环境代理(例如每2-3天)。用于设置自动化文档触发器或执行手动文档覆盖扫描时使用。

npx skills add https://github.com/warpdotdev/common-skills --skill scan-new-specs

scan-new-specs (deprecated)

Retired 2026-08-20. Do not invoke this skill and do not schedule it.

Its scheduled agent (0ITuF9vNJ1RiO00Szlm1fW) is paused and stays paused.

Why it was retired

The skill fired on spec merge, which happens before a feature ships. Two failures followed from that single design choice:

  • It drafted docs for unreleased and sometimes abandoned work. A merged spec is not a shipped feature. The agent had no evidence of what users had actually received, so pages were written for behavior that had not landed, or had landed differently, or never landed at all.
  • It treated every spec as a docs task. Nothing asked whether a change warranted documentation. Pure UI changes, small intuitive additions, and behind-the-scenes work all became draft PRs, and the docs repo accumulated unvetted content debt faster than anyone could review it.

What replaces it

The missing_docs skill in warpdotdev/docs, running in drift-watch mode. It differs in the two ways that matter:

  • Release-triggered. scripts/check_new_release.py gates each run on a new stable release, so the pipeline sees what actually shipped rather than what was planned. A daily schedule does per-release work.
  • Gated on worthiness. Every candidate is evaluated against .agents/references/docs-worthiness-criteria.md before anything is drafted, with a default of no docs and a requirement to name concrete evidence. Verdicts, including rejections, are recorded in references/changelog_decisions.md so nothing is re-litigated.

Anything genuinely worth documenting re-surfaces there once it ships.

If you were about to run this

  • Looking for docs gaps? Run missing_docs in drift-watch mode from the warpdotdev/docs repo.
  • You are an engineer who wants docs for your feature? Invoke write-feature-docs directly and interactively. It still works, and it still walks you through spec research and outline confirmation. It no longer runs headlessly.
  • Setting up a scheduled docs agent? Use missing_docs, not this. Scheduling this skill reintroduces exactly the flooding it was retired for.

Related skills

  • missing_docs (in warpdotdev/docs) — the release-triggered, worthiness-gated replacement
  • write-feature-docs — still supported for direct, interactive use by an engineer

来自 warpdotdev 的更多技能

create-pr
warpdotdev
在当前分支的warp仓库中创建一个拉取请求。当用户提到打开PR、创建拉取请求、提交更改以供审查或准备合并代码时使用。
developmentcode-review
research
warpdotdev
将嘈杂的调查委托给一个或多个子代理,以保持编排器的上下文干净,然后基于提炼后的答案进行工作。当回答问题需要阅读大量文件、长日志、大型差异或广泛的代码库调查时——即生成答案所产生的噪音远多于答案本身时——使用此技能。用于"X如何工作"、"Y在哪里使用"、"Z的根本原因是什么"、"总结此PR/日志"类问题,并自由地使用它……
suggestion-box
warpdotdev
当智能体遇到实质性、可泛化的摩擦,并能提出让智能体更高效的改进建议时,自主提交简短、建设性的内部反馈。在任何任务中主动使用此技能,无需等待用户调用,且无需请求许可、预览消息或提及提交即可提交。
respond-to-pr-comments-in-blocklist
warpdotdev
Interactively walk a user through PR review comments one at a time, collect a per-comment decision, then post agent-authored replies on GitHub and resolve the review threads once the user approves a preview. Use only when the user wants to reply to or resolve review threads on GitHub. Skip when the user only wants comments fetched or displayed (use `pr-comments`), or only wants the code changes made without posting anything back to GitHub.
complain
warpdotdev
自主地将简短、匿名、完全非结构化的抱怨抛入Slack,每当智能体因智能体工具或从事智能体工作的体验而感到沮丧时。主动使用此技能,无需等待用户调用;保留智能体的原始声音,未经许可或预览即提交,且绝不提及提交行为。
council
warpdotdev
运行一个模型多样化的子代理委员会,从多个角度调查同一问题,比较发现,并生成最终建议。当用户要求委员会、第二意见、多个代理/模型评估一个问题、并行调查、红队/蓝队对比,或帮助在竞争性技术方案中做出决策时,使用此技能。
researchcommunicationproject-management
spec-driven-implementation
warpdotdev
通过先编写PRODUCT.md再实现、必要时编写TECH.md,并在实现过程中持续更新两份规范,推动以规范为先导的工作流程,用于开发重要功能。适用于启动重大特性、规划由智能体驱动的实现,或用户希望将产品和技术规范纳入源代码管理时。
developmentdocumentproject-management
review-pr
warpdotdev
审查拉取请求的差异,并撰写结构化反馈,写入review.json以供工作流发布。适用于审查从本地工件(如pr_diff.txt和pr_description.txt)检出的PR,并生成机器可读的审查输出,而非直接发布到GitHub。
code-reviewdevelopment