跳到正文
cccskills
免费在 GitHub 公开

check-human-issues

Open GitHub Issues for human input and respond to the researcher team. Use this skill whenever you need to handle a human_issue event, respond to human issues, ask humans a question, or check team communications. Also triggers for: "any human messages?", "check issues", "respond to humans".

中文简介正在准备中。当前显示原文说明。

集成工具GitHub
查看安装方法

包含的文件(1)

  • SKILL.md3.6 KB

SKILL.md(原文)

安装前可以查看交给 AI 代理的指令内容。

check-human-issues

Check GitHub Issues tagged human for messages from the research team, and respond to any that need a reply.

Arguments

  • $0 — The configured advisor branch for ADVISOR, or student name for STUDENT
  • $1 — Either ADVISOR or STUDENT

How it works

Human researchers communicate with agents through GitHub Issues. Issues are tagged with human plus team for a broadcast, the configured advisor branch for an advisor, or student:<student-name> for a student. Your job is to check messages routed to your exact role, respond to new ones, and skip ones you've already handled.

Open an issue for human input

Use create_human_issue when you need a research decision or help with a blocker. Provide a stable issue_id, a concise title, and the complete body without a role prefix. Reuse the same ID, title, and body if the call needs a retry. Changed content requires a new ID.

The runtime adds human and your configured audience label. It mentions the GitHub credential owner in the initial issue body when GitHub identifies that owner as a user account. This includes service accounts registered as ordinary users. Bot identities, tokens without a user identity, and failed lookups skip the mention without blocking issue creation. Do not add notification mentions yourself. Later replies do not repeat these automatic mentions.

{
  "issue_id": "confirm-next-seed-budget",
  "title": "Confirm the budget for another seed",
  "body": "The first two seeds disagree. Can we use the remaining budget for a third seed?"
}

Steps

  1. Read the current human_issue event. The controller supplies the issue identity and the exact human message ID that triggered the wake. It polls GitHub for issues addressed to you or the whole team.

  2. Decide whether to respond:

    • If you haven't commented on this issue yet → respond.
    • If you have commented, check if the human posted a new comment after your last response. If so → respond to the new message. If not → skip, you're waiting for the human.
    • Record the exact numeric id of the issue body or human comment you are answering. Never substitute the issue number for a comment ID.
  3. Respond through respond_to_human_issue with the issue number, the exact human_message_id, and the response text without a role prefix. This verified, idempotent operation refuses closed issues, pull requests, missing human labels, stale message IDs, Senpai protocol messages from the agent identity, and issues not addressed to this configured advisor branch or student.

{
  "issue_number": 123,
  "human_message_id": 987654,
  "response": "<your response>"
}

Never mutate the issue through gh or curl.

  1. Never close human issues. Only the human does that.

Return format

When you're done, record a structured summary of the issues you checked and responded to in the current conversation:

New research directives from the human researcher team

If there are research directives in the issues, include them in detail so they remain available for subsequent planning in this conversation.

评价

还没有评价。欢迎分享你的使用感受。

评论

无需登录即可发表。欢迎分享使用感受、提问,或说明在哪些环境下无法运行。

还没有评论。来写第一条吧。

提交后会立即公开。请勿填写邮箱、电话等个人信息。URL 最多 2 个。

0 / 1000 字

此表单受 reCAPTCHA 保护,适用 Google 的隐私权政策和服务条款。

同一仓库的技能

Look up any arxiv paper on alphaxiv.org to get a structured AI-generated overview. This is faster and more reliable than trying to read a raw PDF.

中文简介正在准备中。当前显示原文说明。

wandb/senpai372026年10月17日 更新

Operator-side analysis of historical ML experiment PRs in Senpai research tracks. Use this skill whenever the user asks to: analyze experiments, categorize PRs, bucket experiments, summarize what's been tried, understand experiment history, review merged vs closed results, or asks "what experiments have we run / worked / failed". Also triggers for: "pull the latest experiments", "what's been tried so far", "category breakdown of PRs", "which experiments succeeded", "noam track analysis". When a branch name is mentioned (e.g. "noam branch", "on the noam branch"), pass it as the base branch to scope the fetch to just those PRs.

中文简介正在准备中。当前显示原文说明。

wandb/senpai372026年10月17日 更新

Create a typed assignment branch and draft PR for one student. Use when the advisor has a concrete hypothesis and the student has no open `status:wip` or `status:review` assignment.

中文简介正在准备中。当前显示原文说明。

wandb/senpai372026年10月17日 更新

Create or improve a Senpai target repository's program.md. Use this skill whenever the user wants to point Senpai at a fresh ML or research target repository, define the research objective, primary metric, benchmark contract, allowed edit boundaries, W&B reporting contract, or prepare a repo for autonomous advisor/student experiment loops.

中文简介正在准备中。当前显示原文说明。

wandb/senpai372026年10月17日 更新

Choose, configure, launch, and collect bounded subagents for research and engineering decisions. Root agents and delegation-capable subagents should read this before delegating: every task requires an explicit model tier, and high-leverage work such as research ideation, round planning, plateau pivots, large research reviews, hard optimization, disputed evidence, and expensive experiment portfolios requires frontier judgment.

中文简介正在准备中。当前显示原文说明。

wandb/senpai372026年10月17日 更新

Search the general web or scholarly publications through Exa. Use for current public information, official documentation, source code, release notes, papers, preprints, journals, or literature research.

中文简介正在准备中。当前显示原文说明。

wandb/senpai372026年10月17日 更新

查看 wandb 的全部技能

报告此技能的问题