Issue Triage Skill

This skill helps triage open GitHub issues in the dotnet/maui repository by:

Initializing a session with current milestones and labels
Loading a batch of issues into memory
Presenting issues ONE AT A TIME for triage decisions
Suggesting milestones based on issue characteristics
Tracking progress through a triage session

Prerequisites

GitHub CLI (gh) must be installed and authenticated:

# Install
# Windows:
winget install --id GitHub.cli

# macOS:
brew install gh

# Linux:
# See https://cli.github.com/manual/installation

# Authenticate (required before first use)
gh auth login

The scripts will check for gh and exit with installation instructions if not found.

When to Use

“Find issues to triage”
“Let’s triage issues”
“Grab me 10 issues to triage”
“Triage Android issues”

Triage Workflow

🚨 CRITICAL: ALWAYS use the skill scripts. NEVER use ad-hoc GitHub API queries.

The scripts have proper filters, exclusions, and milestone logic built-in. Don’t bypass them.

Step 1: Initialize Session

Start by initializing a session to load current milestones and labels:

pwsh .github/skills/issue-triage/scripts/init-triage-session.ps1

What this does:

Fetches current milestones (SR4, SR5, etc.) from GitHub API
Displays servicing milestones for reference during triage
Creates session file to track progress

Step 2: Load Issues Into Memory

MANDATORY: Use query-issues.ps1 - it has the right filters!

pwsh .github/skills/issue-triage/scripts/query-issues.ps1 -Limit 50 -OutputFormat triage

What this does:

Queries GitHub with exclusion filters: -label:s/needs-info -label:s/needs-repro -label:area-blazor -label:s/try-latest-version -label:s/move-to-vs-feedback
Returns issues without milestones
Stores results for one-at-a-time presentation

DON’T:

❌ Use github-mcp-server-list_issues directly
❌ Use github-mcp-server-search_issues without the same filters
❌ Try to replicate the logic yourself - use the script!

Step 3: Present ONE Issue at a Time

IMPORTANT: When user asks to triage, present only ONE issue at a time in this format:

## Issue #XXXXX

**[Title]**

🔗 [URL]

| Field | Value |
|-------|-------|
| **Author** | username (Syncfusion if applicable) |
| **Platform** | platform |
| **Area** | area |
| **Labels** | labels |
| **Linked PR** | PR info with milestone if available |
| **Regression** | Yes/No |
| **Comments** | count |

**Comment Summary** (if any):
- [Author] Comment preview...

**My Suggestion**: `Milestone` - Reason (based on init session output)

---

What would you like to do with this issue?

Step 4: Wait for User Decision

Wait for user to say:

A milestone name (e.g., “Backlog”, ”.NET 10 SR5”, ”.NET 10 Servicing”)
“yes” to accept suggestion
“skip” or “next” to move on without changes
Specific instructions (e.g., “next SR and add regression label”)

Step 5: Move to Next Issue

After user decision, automatically present the NEXT issue.

Step 6: When Batch is Empty

🚨 CRITICAL: When you run out of issues, AUTOMATICALLY reload more issues.

# Run query again to load next batch
pwsh .github/skills/issue-triage/scripts/query-issues.ps1 -Limit 50 -Skip <current_count> -OutputFormat triage

DO NOT:

❌ Stop and ask “Load more?”
❌ Say “No more issues found”
❌ Use different GitHub queries

DO:

✅ Automatically run query-issues.ps1 again with -Skip parameter
✅ Continue presenting issues one at a time
✅ Only stop when query returns zero issues

Script Parameters

query-issues.ps1

Parameter	Values	Default	Description
`-Platform`	android, ios, windows, maccatalyst, all	all	Filter by platform
`-Area`	Any area label (e.g., collectionview, shell)	""	Filter by area
`-Limit`	1-1000	50	Maximum issues to fetch
`-Skip`	0+	0	Skip first N issues (for pagination)
`-OutputFormat`	table, json, markdown, triage	table	Output format
`-RequireAreaLabel`	switch	false	Only return issues with area-* labels
`-SkipDetails`	switch	false	Skip fetching PRs/comments (faster)

Milestone Suggestion Logic

🚨 CRITICAL: ALWAYS use actual milestone names from init-triage-session.ps1 output. NEVER guess or assume milestone names.

The skill dynamically queries current milestones from dotnet/maui at session initialization. Milestone names change frequently (e.g., SR4, SR5, SR6), so always reference the session output when suggesting milestones.

Suggestion Guidelines

Condition	Suggested Milestone	Reason
Linked PR has milestone	PR’s milestone	”PR already has milestone”
Has `i/regression` label	Highest numbered SR milestone	”Regression - needs servicing”
Has open linked PR	Current servicing milestone	”Has open PR”
Default	Backlog	”No PR, not a regression”

Example Session Output:

Servicing Releases:
  - .NET 9 Servicing [246 open]
  - .NET 10 Servicing [213 open]
  - .NET 10 SR5 [55 open]         ← Use this for .NET 10 regressions
  - .NET 10.0 SR4 [103 open]

Other:
  - .NET 11 Planning [167 open]
  - .NET 11.0-preview1 [8 open]

Backlog:
  - Backlog [3037 open]

How to suggest milestones:

✅ “Assign to .NET 10 SR5” (from session output)
❌ “Assign to .NET 10 SR2” (guessing, might not exist)
❌ “Assign to current SR” (ambiguous, multiple active)

Applying Triage Decisions

# Set milestone only
gh issue edit ISSUE_NUMBER --repo dotnet/maui --milestone "Backlog"

# Set milestone and add labels  
gh issue edit ISSUE_NUMBER --repo dotnet/maui --milestone "MILESTONE_NAME" --add-label "i/regression"

# Set milestone on both issue AND linked PR
gh issue edit ISSUE_NUMBER --repo dotnet/maui --milestone "MILESTONE_NAME"
gh pr edit PR_NUMBER --repo dotnet/maui --milestone "MILESTONE_NAME"

Label Quick Reference

Regression Labels:

i/regression - Confirmed regression
regressed-in-10.0.0 - Specific version

Priority Labels:

p/0 - Critical
p/1 - High
p/2 - Medium
p/3 - Low

iOS 26 / macOS 26:

version/iOS-26 - iOS 26 specific issue

Common Mistakes to Avoid

Mistake	Why It’s Wrong	Correct Approach
❌ Using `github-mcp-server-list_issues` directly	Missing exclusion filters (needs-info, needs-repro, etc.)	✅ Use `query-issues.ps1` script
❌ Stopping when batch is empty	There are likely more issues available	✅ Automatically run `query-issues.ps1 -Skip N`
❌ Suggesting milestone names like “SR2”	Milestone doesn’t exist, based on assumptions	✅ Use actual milestone names from `init-triage-session.ps1` output
❌ Asking “Load more?” when out of issues	Creates unnecessary interruption	✅ Just load more automatically
❌ Using ad-hoc API queries with custom filters	Likely to miss or include wrong issues	✅ Trust the skill’s scripts - they have the right logic

Session Tracking (Optional)

# Record triaged issue
pwsh .github/skills/issue-triage/scripts/record-triage.ps1 -IssueNumber 33272 -Milestone "Backlog"

# View session stats
cat CustomAgentLogsTmp/Triage/triage-*.json | jq '.Stats'

issue-triage

快捷安装