菜单
Recurring check-in on a stack of dependent PRs — rebase children onto updated parents, surface CI state, flag anything waiting on a human-only gate, and report only what changed since last time. Meant to be handed to a time-based loop, not run once.
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。
基于 SOC 职业分类
| name | stacked-pr-loop |
| description | Recurring check-in on a stack of dependent PRs — rebase children onto updated parents, surface CI state, flag anything waiting on a human-only gate, and report only what changed since last time. Meant to be handed to a time-based loop, not run once. |
| disable-model-invocation | true |
A stack is a chain of PRs where each depends on the one below it merging first (main <- A <- B <- C). Left unattended, a stack rots two ways: a lower PR updates and the ones above it go stale, or CI needs a nudge (a manual trigger comment, a required check that didn't fire) and nobody's watching. This is one check-in pass, meant to run on an interval, not a one-shot fix.
List the PRs in the chain and their actual base branches (not assumed order — read each PR's base). Note which ones are merged, open, or closed-unmerged; a closed-unmerged PR downstream of it needs re-basing onto whatever the next real base is.
Done when you have the true current shape of the stack, bottom to top.
For each open PR whose base has moved since the PR's branch was last updated (the base PR merged, or main advanced and this is the bottom of the stack): rebase the branch onto the current base and push. Do this bottom-up — rebasing a PR before the one below it is settled just means redoing the work.
Done when every open PR in the stack is rebased onto its actual current base, or you've confirmed it already was.
For each PR, get the live per-check state against its current head (a rebase invalidates prior runs). If a check is required but hasn't run, that's a red flag, not a pass — don't read absence as success.
Some CI systems require a maintainer-posted trigger comment for a check to run at all (a bot/fork-safety gate). If a PR's CI never fired and the repo uses one of these gates, name it as the specific thing blocking progress — don't just report "CI hasn't run." This is a gate only a human can clear; the loop's job is surfacing that it's waiting, not working around it.
Compare against the last pass. Report:
main. Merge bottom-up only if you have standing authorization to merge without asking each time; otherwise surface it and stop.Stop condition for the loop as a whole: the entire stack has merged, or a full pass produces no state changes from the previous one.
Recurring maintenance pass over automated dependency-bump PRs (Dependabot, Renovate, or similar) — rebase, relock, verify, and report what's ready to merge. Meant to be handed to a time-based loop or schedule, not run once.
Post-deploy verification loop — poll a rollout until every instance is on the new version and healthy, then run one smoke check against a real user path. Meant to be handed to a time-based loop with a timeout, not polled by hand.
Reference for designing agent loops — cycles of work that repeat until a stop condition is met. Covers the four loop shapes, writing completion criteria, carrying state and isolating work across cycles, and what running unattended still leaves on you.
Pick the skill or flow that fits your situation. A router over the skills in this repo.
Review the changes since a fixed point (commit, branch, tag, or merge-base) along two axes — Standards (does the code follow this repo's documented coding standards?) and Spec (does the code match what the originating issue/PRD asked for?). Runs both reviews in parallel sub-agents and reports them side by side. Use when the user wants to review a branch, a PR, work-in-progress changes, or asks to "review since X".
Shared vocabulary and principles for designing deep modules. Use when the user wants to design or refine a module's interface, hunt for deepening opportunities, decide where a seam belongs, make code more testable or AI-navigable, or when another skill needs the deep-module vocabulary.