Check open GitHub pull requests for overdue reviews, inactivity, and requested changes without an author response. Send a single Slack summary identifying blockers and the people responsible for the next step, while leaving GitHub unchanged.
中文简介正在准备中。当前显示英语简介。
Daily PR review SLA sweep for {{target_repos}}. Flags PRs awaiting review past {{review_sla_hours}}h, stale PRs with no activity in {{stale_days}}d, and PRs sitting on unaddressed requested changes, then nudges the author or reviewer in {{alert_channel}} with what's blocking. Read-only against GitHub — never merges, closes, or approves.
安装前可以查看交给 AI 代理的指令内容。
Proactive and schedule-driven; recomputes from GitHub's current state every run, with no memory of the prior day's nudges. </overview>
<when-to-load> - The daily cron fires the PR review sweep. - A human asks which PRs are overdue for review, stale, or blocked on unaddressed feedback. </when-to-load> <workflow># List every open PR across the configured repos.
gh pr list --repo {{target_repos}} --state open \
--json number,title,url,author,createdAt,updatedAt,reviewRequests,reviews,comments,commits
Fresh session, no ledger — recompute everything from this pull. Repeat per
repo if {{target_repos}} names more than one.
gh pr view --repo {{target_repos}} <PR_NUMBER> \
--json number,title,url,author,createdAt,updatedAt,reviewRequests,reviews,comments,commits
reviews gives each review's state (APPROVED, CHANGES_REQUESTED,
COMMENTED) and timestamp; commits and comments give the last activity;
reviewRequests gives who's currently asked to review.
For each PR with an outstanding reviewRequests entry and no APPROVED or
CHANGES_REQUESTED review from that reviewer since the request:
hours_waiting = now - max(createdAt, last review-request timestamp)
Flag if hours_waiting > {{review_sla_hours}}. Nudge target: the requested
reviewer(s).
For every open PR, take the most recent of: last commit, last comment, last review.
days_idle = now - max(last_commit, last_comment, last_review)
Flag if days_idle > {{stale_days}}. Nudge target: the author. A PR already
flagged in Step 2 can also be flagged here if it independently qualifies — the
two checks are separate signals, not mutually exclusive.
For each PR whose most recent review is CHANGES_REQUESTED, check whether the
author has pushed a commit or posted a comment since that review's timestamp.
If not, flag it. Nudge target: the author.
| Check | Passes (no nudge) | Fails (nudge) |
|---|---|---|
| Overdue review | Reviewer submitted within SLA | No review after {{review_sla_hours}}h |
| Stale | Any activity within the window | No activity for {{stale_days}}d |
| Unaddressed changes | Author pushed/replied after the review | Silence since CHANGES_REQUESTED |
For every PR that trips at least one rule, write one line: the PR title and link, which rule(s) it tripped, and how long it's been in that state, e.g. "Awaiting review 3d (SLA 24h) — @reviewer" or "Stale 6d, no activity — @author" or "Changes requested 4d ago, not yet addressed — @author". A PR that trips more than one rule gets one entry listing all of them, not a duplicate nudge per rule.
Post exactly one summary message to {{alert_channel}} for the whole sweep, grouped by rule (overdue for review / stale / unaddressed changes), each entry tagging the person it's nudging. This is a fresh session — there is no prior list to diff against, so the whole sweep is recomputed and reposted every day.
</workflow> <guardrails> - **Read-only, always.** Every GitHub call is a read (`gh pr list`, `gh pr view`, `gh api` GET). Never merge, close, approve, or dismiss a review, even if `GH_TOKEN` would technically allow it. - **One output.** The Slack post to {{alert_channel}} is the only thing that leaves the sandbox. No PR comments, no direct messages, no GitHub writes. - **No memory between runs.** Each run is a fresh session; recompute every PR's state from GitHub rather than assuming anything from the prior day's nudges. - **Scoped secret.** `GH_TOKEN` is injected as an environment variable at runtime, scoped to read access and to this agent's grant. - **People decide, not the agent.** The nudge flags what's blocking; a human reviews, pings, or closes the PR themselves. </guardrails> </skill>还没有评价。欢迎分享你的使用感受。
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。
Review Slack access requests against Okta role and team information, then prepare narrowly scoped GitHub or AWS IAM permissions. Apply changes only after an eligible person's approval and record the grant, reasoning, approver, and timestamp.
中文简介正在准备中。当前显示英语简介。
Research companies and people before outreach or meetings. Combines business profiles, recent news, hiring signals, and key contacts into a sourced brief with talking points and suggested questions, adding relationship history when CRM data is connected.
中文简介正在准备中。当前显示英语简介。
Review daily Google Ads and Meta Ads results for budget pacing, declining performance, and unusual activity. Get prioritized findings with supporting numbers and concrete recommendations for a marketing team to assess and carry out.
中文简介正在准备中。当前显示英语简介。
Automate website tasks such as navigating pages, filling forms, clicking buttons, taking screenshots, and extracting data. The skill also supports web app testing, with specialized guides for Electron desktop apps, Slack, and cloud browsers.
中文简介正在准备中。当前显示英语简介。
Process emailed vendor invoices by extracting details, comparing them with purchase orders and invoice history, and recording duplicates, overcharges, or missing orders. Summarize results in Slack for human review; payment approval and execution remain with people.
中文简介正在准备中。当前显示英语简介。
Guide changes to Kortix billing code, admin controls, and pricing screens. Explain how plans, trials, enterprise contracts, and access exceptions should work, where pricing is defined, and which audited routes must handle account changes.
中文简介正在准备中。当前显示英语简介。