notque/vexjoy-agent / skills/process/workflow
workflow
VerifiedStructured work: multi-phase tasks, feature builds, planning, objective loops, hill climbing.
references/pr-fix-loop/references/review-checklist.md
View on GitHub# PR Fix Loop: Review Checklist
Load in Phase 1. Every applicable item is tested or marked N/A with a reason. A domain overlay adds its own items to the end of each list.
## Comment collection
Comments live on the PR's base repo. Resolve it from the PR URL, so the commands work from a fork checkout too:
```bash
REPO=$(gh pr view "$PR" --json url -q .url | cut -d/ -f4-5) # owner/name of the base repo
gh api --paginate "repos/$REPO/pulls/$PR/reviews" # bot overview and review bodies
gh api --paginate "repos/$REPO/pulls/$PR/comments" # inline comments, with commit_id and path:line
gh api --paginate "repos/$REPO/issues/$PR/comments" # conversation comments
```
On a GitHub Enterprise Server host, add `--hostname <host>` (the third field of the PR URL).
Write one file per PR, for example `pr-<n>-comments.md`, with each comment's source, `commit_id`, `path:line`, and status: open, fixed in `<sha>`, or invalid (reason in a local file).
A comment made on an older commit may already be fixed. Check it against the current head before acting.
## Render check
Judge what users see, not the config that produces it.
**Any change**
Run the changed output the way a user meets it: CLI output, API responses, rendered docs, generated config, or UI. Record the before and after output for the same input.
**Observability changes (dashboards, alerts, metrics)**
Dashboards:
1. Load the dashboard in the real UI against live data.
2. Screenshot every panel.
3. Compare each panel's value to an independent source, such as a direct query or the system's own counter.
4. Record per panel: shown value, independent value, match or mismatch.
Alerts:
1. Run the rule unit-test tool (for Prometheus, `promtool test rules`) against real label sets, including extra labels the series carries in production.
2. Evaluate each expression on the last 24h of live data.
3. Per alert, record: would it fire now, and would that be noise.
## Edge cases
**Generic (every PR)**
| # | Case | What to check |
|---|------|---------------|
| G1 | Upgrade path | Existing installs and configs keep working after the change; defaults did not shift silently |
| G2 | Other consumers | Other users of the same module, chart, or API see no unintended change |
| G3 | Error paths | Failure and empty-input paths behave as before or better |
| G4 | Alternate paths | Every code or data path that reaches the change behaves the same; paused paths count as live |
**Observability changes**
| # | Case | What to check |
|---|------|---------------|
| O1 | Rollout and restart | Terminated pods still inside rate windows; counter resets on restart or OOM |
| O2 | Multiple installs | Two or more installs in one metrics backend; selectors and aggregation keep them apart |
| O3 | Absent series | Series missing at start; counters that only appear after the first failure |
| O4 | Failover-level failure | Every failover level failing at once, not just the primary |
| O5 | Shared charts | Other teams' installs use the same chart; defaults and value changes affect them |
| O6 | Peak shape | Message size and batch shape at peak volume, including volume from paused paths |
## Re-review rules
- Re-verify each claim in the PR body and in review comments against live data, read-only.
- Use paused-path data as the baseline, not today's reduced traffic.
- Quote the query and the number for every claim you confirm or reject.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| planning-and-task-breakdownaddyosmani/agent-skillsBreaks work into ordered tasks. Use when you have a spec or clear requirements and need to break work into implementable tasks. Use when a task feels too large to start, when you need to estimate scope, or when parallel work is possible. | Stars: 119k | Updated | Claude CodeCodex | Verified | Install |
| screenpipethedotmack/claude-memSearch screen and audio history recorded by Screenpipe and optionally save selected context into claude-mem. Use when the user asks to recall desktop activity, a meeting, or something seen in another app through Screenpipe. | Stars: 118k | Updated | Claude CodeCodex | Verified | Install |
| contact-widgetnexu-io/open-designSelf-contained floating chat widget with welcome screen, social links, meeting button, and message input. Single HTML file, zero dependencies. | Stars: 108k | Updated | Claude CodeCodex | Verified | Install |
| meeting-notesnexu-io/open-designMeeting notes page — title bar with attendees, agenda checklist, decisions block, action items table with owners + dates, and a "next meeting" footer. Use when the brief mentions "meeting notes", "minutes", "1:1 notes", "all-hands recap", or "会议纪要". | Stars: 108k | Updated | Claude CodeCodex | Verified | Install |