maintainer-briefing

name	maintainer-briefing
description	Generate a prioritized maintainer briefing from current GitHub repository state, including issue/PR dependencies, CI, review blockers, regressions, and drift.
user_invocable	true

Maintainer Briefing

Allocate maintainer attention; do not perform detailed code review or invent merge recommendations.

Resolve the repository argument or use:

gh repo view --json nameWithOwner -q '.nameWithOwner'

Fetch open pull requests and issues with gh, including review decisions, unresolved threads, labels, checks, timestamps, authors, and size. Fetch details only for candidates likely to appear in the report.

Dependency map

Build an issue/PR map from closing references, linked issues, comments, commit messages, and overlapping symbols when available. Detect:

• issues with open, merged, or superseded fixes;
• PRs whose premise changed because another PR merged first;
• reopened regressions or partial fixes;
• stale file and line references after ports or large insertions;
• contradictory issues or PRs touching the same behavior;
• base-branch failures affecting multiple pull requests.

Do not infer a dependency from similar titles alone.

Prioritization

Rank by:

1. security or data-loss risk;
2. contributor blocked on a maintainer decision;
3. regression or broken release/CI;
4. user already engaged and context still warm;
5. small, passing goodwill wins;
6. stale cleanup.

Estimate effort from actual scope and decision complexity, not line count alone. Deduplicate items across sections.

Report

Keep the report compact:

Critical

At most three security, regression, release, or contributor-blocking items.

Decisions Needed

Items requiring maintainer product/API direction rather than code review.

Easy Wins

Small, passing, low-risk work with the exact next action and effort estimate.

Deep Focus

At most three high-leverage reviews or investigations, each under 30 minutes unless unavoidable.

Dependency and Drift

Superseded fixes, reopened issues, stale citations, contradictory premises, and safe close/ping candidates.

Snapshot

Open PR/issue counts, items waiting on the maintainer, CI failures, and backlog distribution.

Close with the single best action today, fastest contributor goodwill win, and what is safe to ignore. Ground every item in fetched data and use #number references consistently.