personal-pr-triage

name	personal-pr-triage
description	Automates the weekly PR triage for Flutter teams, following the standards in docs/triage/README.md.

PR Triage Skill

This skill automates the weekly PR triage for Flutter teams, following the standards in docs/triage/README.md.

Workflow

1. Identify Team and Retrieve PR Lists

• Locate the team's PR queries in docs/triage/README.md.
• Retrieve ALL open, non-draft PRs for the primary repository and any owned packages.
• For large volumes (e.g., >20 PRs), first fetch numbers to a file to avoid API timeouts: gh pr list --search "[QUERY] -is:draft" --limit 200 --json number --jq '.[].number' > pr_numbers.txt

2. Fetch Detailed Audit Data

For each PR number, fetch comprehensive JSON. For large volumes, use a script to fetch individually or in smaller batches to avoid 502 errors.

• Required fields: number,title,author,authorAssociation,isDraft,reviews,comments,reviewRequests,statusCheckRollup,updatedAt,reviewDecision,labels,url,commits.
• For smaller sets, a loop is fine:

echo "[" > audit_data.json
first=true
for num in $(cat pr_numbers.txt); do
  if [ "$first" = true ]; then first=false; else echo "," >> audit_data.json; fi
  gh pr view $num --json number,title,author,authorAssociation,isDraft,reviews,comments,reviewRequests,statusCheckRollup,updatedAt,reviewDecision,labels,url,commits >> audit_data.json
done
echo "]" >> audit_data.json

• For larger sets (>50 PRs), prefer a Python fetcher script.

3. Generate the Report

Run the bundled triage script to process the data and generate triage_report_pr_[TEAM].md.

python3 .agents/skills/personal-pr-triage/scripts/generate_triage_report.py [TEAM_NAME]

⚠️ Critical Verification Mandate (Zero-Trust)

1. Succinct Formatting & Zero Redundancy: Keep reports extremely concise. Omit redundant bullet points where the section header already explains the action (e.g., under Needs Initial Reviewer or Fix Review Feedback, do not repeat **To-do**: Assign reviewer or **To-do**: Author to address comments). Merge status and actions into a single bullet point per PR. Truncate long lists of failing CI checks (>3 checks) to keep reports scannable.
2. Action Precision & Grouping:
   • Author Actions Grouping: Group Author Actions (Action Required) into specific granular categories based on the action required:
     • 🛑 Action Required: Fix Review Feedback (Changes requested by reviewers)
     • 📝 Action Required: Sign CLA (Failing CLA check cla/google)
     • ⚠️ Action Required: Fix CI (Failing functional CI checks, excluding CLA and freeze)
     • ❄️ Action Required: Address Code Freeze Check (Failing code freeze checks)
     • 💬 Action Required: Ping Author (Idle 14+ Days) (Idle, needs a first ping)
     • 🔒 Action Required: Close PR (Idle 28+ Days) (Idle >28 days, needs closure)
     • ⏳ Waiting on Author (Already Pinged) (Idle, already pinged)
   • Google testing fail -> "Googler to assist".
   • Stale + last comment is member ping -> Omit redundant todo bullets.
   • Approved + Blocked CI -> "To-do: Nudge author (approved by @user)".
3. Stale calculation: If the last activity was a member pinging the author, calculate the "stale days" from that ping.
4. CI Requirement for Landing: Only list PRs in "Ready to Land" if CI is completely green.
5. Deterministic Logic: Always use the provided Python script to ensure consistency across triage sessions.
6. Embedded Link & Heading Standards: Heading level 3 (###) must be the highest-level heading in the report (used for the title and main groups). Subgroups must use heading level 4 (####) with embedded links for section counts: #### [TITLE] ([COUNT](LINK)).

🛠 Useful Scripts

Fetch detailed state:

gh pr view [NUMBER] --json number,title,author,authorAssociation,isDraft,reviews,comments,reviewRequests,statusCheckRollup,updatedAt,reviewDecision,labels,url,commits