用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/arthur-debert/release --skill handoff命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
Drive a PR to ready-for-human-merge in any repo managed by `arthur-debert/release`. The loop is state-machine-driven: `release-core pr status` reports one lifecycle state plus the single next action; do that action, re-read, repeat — open as a draft, request the required reviews, wait in-turn, triage and resolve threads — and `release-core pr wait` flips draft→ready itself once the engine reaches READY, then stop (the human merges). Use when opening a PR, checking where a PR stands, waiting on or triaging review feedback, or driving a PR toward merge-readiness. Triggered by: `gh pr create`, 'check PR status', 'where does this PR stand', requesting a review, or processing review comments.
Escalate infrastructure friction (workflow failures, ruleset misbehaviors, broken policy templates, helper-script bugs) from a consumer repo to the `arthur-debert/release` repo. Files a GitHub issue at release with auto-collected reproduction context (repo, branch, PR, recent workflow run, workaround applied). Searches recent release issues first and comments on a matching one rather than filing a duplicate — the comment count is the signal of recurrence. Use ONLY for infra issues that the consumer cannot fix in place; do not use for code-quality nits, project-specific test failures, or anything fixable inside the consumer repo. Triggered by: 'this workflow fails in a way I can't fix here', 'the release-loop policy is misbehaving', 'an infra script errors', 'the cloud env is missing something it should provide'.
Measure whether automated PR reviews (Copilot / Gemini / CodeRabbit) actually earn their cost on a given repo, so review policy is set by evidence instead of gut. Runs a repo-agnostic pipeline over the repo's own merged-PR history: a cheap mechanical tier (extract → enrich → finalize → summarize, zero agents) answers latency / volume / convergence, and an optional LLM-judge tier (a `release` Workflow fan-out + aggregate → converge) scores each bot comment for value, hallucination, novelty and re-request convergence. Emits a machine-readable summary.json. Use when: 'audit our PR reviews', 'are the review bots worth it', 'should we keep CodeRabbit / Gemini', 'is re-request-on-push paying off', 'one reviewer or several?'.
基于 SOC 职业分类
正在显示 SKILL.md
| argument-hint | What will the next session be used for? |
| description | Compact the current conversation into a handoff document for another agent to pick up. |
| metadata | {"github-path":"skills/productivity/handoff","github-ref":"refs/heads/main","github-repo":"https://github.com/mattpocock/skills","github-tree-sha":"7bbb60d64c911617cee7f71b2ba9e24364faa14e"} |
| name | handoff |
Write a handoff document summarising the current conversation so a fresh agent can continue the work. Save to the temporary directory of the user's OS - not the current workspace.
Include a "suggested skills" section in the document, which suggests skills that the agent should invoke.
Do not duplicate content already captured in other artifacts (PRDs, plans, ADRs, issues, commits, diffs). Reference them by path or URL instead.
Redact any sensitive information, such as API keys, passwords, or personally identifiable information.
If the user passed arguments, treat them as a description of what the next session will focus on and tailor the doc accordingly.
