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/max-sixty/tend --skill weekly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。
正在显示 SKILL.md
| name | weekly |
| description | Weekly maintenance — reviews dependency PRs and runs any repo-specific weekly tasks defined in running-tend. |
| metadata | {"internal":true} |
Load /tend-ci-runner:running-in-ci first — it contains CI security rules, review/comment formatting, and polling conventions. This skill posts approvals and comments on PRs, so those rules apply. running-in-ci will also load the repo's running-tend overlay if one exists; keep the loaded content in mind for Step 3.
gh pr list --state open --limit 200 --json number,title,author,labels \
--jq '.[] | select(.author.login == "dependabot[bot]" or .author.login == "renovate[bot]" or (.labels | any(.name == "dependencies")))'
If no dependency PRs are open, note "0 dependency PRs to process" and continue to Step 3 — do not exit; repo-specific weekly tasks may still be due.
Check CI status: gh pr checks <number>
If CI is passing, review the diff for breaking changes (major version bumps, API changes, deprecation warnings)
If the update is safe (patch/minor with green CI), check whether the bot has already approved this commit before approving — a dependabot PR open across multiple weekly runs (or already approved by tend-review on creation) would otherwise accumulate redundant approvals on the same commit_id:
uv run --script "${CLAUDE_PLUGIN_ROOT}/scripts/bot_review_state.py" \
prepare-approval <number>
If already_approved is true, this PR is done — move to the next one.
Otherwise compose $TMPDIR/review-body.md. Give the
reviewer the context for the approval: the upgrade's scope and the evidence
relevant to its safety. Keep it concise and omit the inspection chronology.
Use a file rather than an inline --body because a
package name written as inline code puts a backtick in a double-quoted
argument, and bash runs the span as a command. Then post, re-reading the sha
from disk, since shell state didn't survive the Write:
# `commit_id` pins the approval to the commit that was checked. Unpinned,
# GitHub anchors it at whatever is live when the POST lands — an approval
# of code nothing checked, on a PR `nightly` rebases on purpose. Read the
# sha first and bail if it isn't there: inlined as `$(cat ...)` a missing
# file substitutes the empty string and the POST still runs, which is the
# unpinned approval this pins against.
CHECKED=$(cat "$TMPDIR/checked-head-<number>") || exit 0
REPO=$(gh repo view --json nameWithOwner --jq '.nameWithOwner')
gh api "repos/$REPO/pulls/<number>/reviews" --method POST \
-f event=APPROVE -f commit_id="$CHECKED" -F body=@"$TMPDIR/review-body.md"
If CI is failing, comment with the failure summary and skip
If a major version bump, comment noting it needs manual review and skip
On either skip path (4 or 5), dismiss an approval that predates the newest rewrite before you leave. Both paths are reachable because a rebase changed something, and neither passes through item 3's guard — so the pre-rewrite approval stays the bot's latest review, re-anchored onto the current head, and the PR still reads as bot-approved while you comment that it isn't mergeable:
uv run --script "${CLAUDE_PLUGIN_ROOT}/scripts/bot_review_state.py" \
dismiss-stale <number> \
"Rebased since this approval; the new head is unreviewed."
A dismissed review reports DISMISSED rather than APPROVED, so the filter stops matching it and a later run re-dismisses nothing.
Perform any weekly maintenance the loaded running-tend overlay defines, following the repo's PR title conventions. If it defines no weekly tasks (or none are due this week), say so in the summary.
Report: dependency PRs processed/approved/skipped (with reasons), and repo-specific weekly tasks completed (or "no repo-specific weekly tasks defined" / "no weekly tasks due").
Tend-specific guidance for tend CI workflows. Adds non-standard workflow inclusion for usage analysis and repo conventions on top of the generic tend-* skills.
Outcome-based analysis of tend's CI behavior — checks whether tend's outputs were accepted or rejected, escalating to session logs only when outcomes look wrong.
Reviews a pull request for code quality and correctness. Use when asked to review a PR or when running as an automated PR reviewer.