github-issue-triage
Use when you have a backlog of unorganized GitHub Issues — bulk-reads, prioritizes, and labels issues at scale using `gh` (the default when installed) or Copilot's built-in GitHub MCP tools when exposed, and proposes assignees
Install
npx skills add https://github.com/drvoss/everything-copilot-cli/tree/main/skills/copilot-exclusive/github-issue-triage
claude plugin marketplace add https://llmmart.ai/marketplace.json && claude plugin install drvoss-everything-copilot-cli@llmmart
git clone https://github.com/drvoss/everything-copilot-cli.git
The skills CLI installs just this skill, for any of its supported agents. Claude Code installs the whole drvoss/everything-copilot-cli collection as a plugin from our marketplace. Git is the plain clone.
Skill manifest
GitHub Issue Triage & Management
Why This is Copilot-Exclusive
GitHub access is built into Copilot CLI with no extra setup: the built-in GitHub MCP server (full
tool set when gh is absent, or when restored with the flags in the note below) or gh, which
the model is expected to use when it is installed. When exposed, the MCP issue tools add sub-methods for comments,
labels, and sub-issues. Either way Copilot reasons over the results directly. What is exclusive is the built-in GitHub MCP server, which needs no token or extension setup; when gh is installed, the model mostly uses gh, which any agent with a shell can also run.
Note: When
ghis on PATH, Copilot CLI omits these issue tools (list_issues,search_issues,issue_read) and the model is expected to useghinstead, e.g.gh issue list --state open --label <label>,gh issue view <n> --comments,gh search issues "<query>". To call the MCP tools, start the session with--add-github-mcp-tool=<name>for every GitHub MCP tool you need (it replaces the default set, so include default ones you still use), or--enable-all-github-mcp-tools, which also enables write tools.
When to Use
- Triaging a backlog of open issues across one or more repositories
- Auto-categorizing issues by content (bug, feature, docs, etc.)
- Finding duplicate or related issues before filing new ones
- Building one queue from both issues and external-contributor PRs
- Linking issues to pull requests for traceability
- Generating weekly issue reports or burndown summaries
Workflow
1. List Open Issues with Filters
With gh on PATH (default), e.g. gh issue list --repo my-org/my-app --state open --label bug --limit 25. When the MCP tools are exposed:
Tool: github-mcp-server-list_issues
owner: "my-org"
repo: "my-app"
state: "OPEN"
labels: ["bug"]
orderBy: "UPDATED_AT"
direction: "DESC"
perPage: 25
2. Search Across Repositories
Find issues matching a keyword across your entire org.
With gh on PATH (default), e.g. gh search issues "memory leak" --owner my-org --state open. When the MCP tools are exposed:
Tool: github-mcp-server-search_issues
query: "memory leak is:open"
owner: "my-org"
3. Read Issue Details and Comments
With gh on PATH (default), e.g. gh issue view 87 --repo my-org/my-app --comments. When the MCP tools are exposed:
Tool: github-mcp-server-issue_read
method: "get"
owner: "my-org"
repo: "my-app"
issue_number: 87
Tool: github-mcp-server-issue_read
method: "get_comments"
owner: "my-org"
repo: "my-app"
issue_number: 87
4. Check Sub-Issues
For tracking epics and parent/child relationships.
With gh on PATH (default), e.g. gh api repos/my-org/my-app/issues/42/sub_issues. When the MCP tools are exposed:
Tool: github-mcp-server-issue_read
method: "get_sub_issues"
owner: "my-org"
repo: "my-app"
issue_number: 42
5. Get Labels for Classification
With gh on PATH (default), e.g. gh issue view 87 --repo my-org/my-app --json labels. When the MCP tools are exposed:
Tool: github-mcp-server-issue_read
method: "get_labels"
owner: "my-org"
repo: "my-app"
issue_number: 87
6. Pull External-Contributor PRs into the Queue
If the inbox includes PRs as well as issues, filter for outside contributors first and then route those PRs through the same triage state model you use for issues.
gh pr list --limit 100 --json number,title,authorAssociation,isDraft,url |
ConvertFrom-Json |
Where-Object {
$_.authorAssociation -in @('CONTRIBUTOR', 'FIRST_TIME_CONTRIBUTOR', 'NONE') -and
-not $_.isDraft
}
Treat OWNER, MEMBER, and COLLABORATOR PRs as normal maintainer review unless your project has
a stronger triage requirement. If a bare reference such as #42 could point to either a PR or an
issue, check the PR first and fall back to the issue only if no PR exists.
When an external PR duplicates existing work, close the loop explicitly: mark it wontfix, link
the authoritative issue or PR, and avoid creating a second parallel thread for the same request.
Examples
Weekly Issue Triage Session
"Show me all open issues in my-org/api-service created in the last 7 days, grouped by label. For unlabeled issues, suggest a category based on the title and body."
Copilot runs gh issue list --repo my-org/api-service --search "created:>=<date>" (when exposed: list_issues with a since filter), reads each issue's labels,
groups them, and uses the issue body to suggest labels for uncategorized ones.
Duplicate Detection
"Before I file a bug about 'WebSocket disconnects on idle timeout', check if there's already an issue about this."
With gh on PATH (default), e.g. gh search issues "WebSocket disconnect idle timeout" --repo my-org/my-app. When the MCP tools are exposed:
Tool: github-mcp-server-search_issues
query: "WebSocket disconnect idle timeout"
owner: "my-org"
repo: "my-app"
Copilot searches, reads the top matches, and tells you whether to file new or comment on an existing issue.
Sprint Planning Report
"List all issues labeled 'sprint-12' in my-org/frontend, show their status, assignee, and whether they have linked PRs."
Copilot lists the issues, checks each for linked PRs via comments and cross-references, and produces a sprint board summary right in the terminal.
Shared Issue + PR Intake
"Show me new outside-contributor PRs and new unlabeled issues from the last week, then tell me which ones need triage first."
Copilot lists issues with gh issue list (or the MCP issue tools when exposed), pulls in external PRs with gh pr list, then routes
the highest-signal items into the same triage queue instead of treating PRs as a separate inbox.
Batch Categorization with SQL Tracking
-- Track triage progress in session database
CREATE TABLE IF NOT EXISTS triage_items (
issue_number INTEGER, repo TEXT, title TEXT,
suggested_label TEXT, status TEXT DEFAULT 'pending'
);
Load issues into the session database, then work through them systematically.
When the Report Is a Vulnerability
A security report must not be triaged in public. If someone opens an issue describing a vulnerability, do not discuss the details in the thread — point the reporter to the repository's private vulnerability reporting (Security → Advisories), then close and lock the public issue. Neither makes its content private: if it contains exploit or other sensitive details, ask a repository admin to delete it — in an organization, an owner may first have to allow issue deletion.
For the advisory itself, the agent prepares evidence and fixes; a maintainer accepts, merges, releases, and publishes:
| Step | Who |
|---|---|
| Verify the claim: who owns the code path, and which release lines are affected | agent, then a maintainer confirms |
| Accept the advisory (the reporter sees this) | human only |
| Fix in the advisory's temporary private fork and review there, not in a public PR | agent may prepare the fix |
| Merge to every affected release line and release each one | agent lists the lines; a maintainer merges and releases |
| Publish the advisory — there is no unpublish | human only |
After publication, follow-up issues, labels, and board cards fall under the repository's normal triage rules.
Do not copy the reproduction or proof of concept onto a project board or a tracking issue: people
with project access are not necessarily advisory collaborators. Link to the advisory instead.
Before publishing, check the credits and decide on a CVE. For reports that came in through private
vulnerability reporting, GitHub adds the reporter as a credited user automatically; for an advisory a
maintainer opened, add the credits explicitly. This mirrors the maintainer flow
modelcontextprotocol/inspector keeps in its repository (documentation under CC-BY-4.0).
Tips
- Parallel issue reads: Call
issue_read(withgh:gh issue view) on multiple issues simultaneously to build a triage report fast. - Use
search_issues(withgh:gh search issues) for cross-repo queries: The search syntax supportsorg:,repo:,label:,assignee:,milestone:, and date ranges. - Combine with session SQL: Store triage results in the session database to track which issues you've reviewed and what actions you took.
- Sub-issues for epic tracking: Use
get_sub_issues(withgh:gh api .../issues/<n>/sub_issues) to check progress on large features broken into subtasks. - Link to PRs: After identifying an issue, ask Copilot to search for PRs that reference it — closing the loop between issues and code changes.
- Pair with
triagefor per-item state: Useworkflow/triagewhen one issue or one external PR needs a durableneeds-info/ready-for-agent/wontfixdecision.
Files (everything-copilot-cli)
-
SKILL.md 9.2 KB
--- name: github-issue-triage license: MIT description: Use when you have a backlog of unorganized GitHub Issues — bulk-reads, prioritizes, and labels issues at scale using `gh` (the default when installed) or Copilot's built-in GitHub MCP tools when exposed, and proposes assignees metadata: category: copilot-exclusive copilot_feature: "Built-in GitHub MCP server (list_issues, search_issues, issue_read)" --- # GitHub Issue Triage & Management ## Why This is Copilot-Exclusive GitHub access is built into Copilot CLI with no extra setup: the built-in GitHub MCP server (full tool set when `gh` is absent, or when restored with the flags in the note below) or `gh`, which the model is expected to use when it is installed. When exposed, the MCP issue tools add sub-methods for comments, labels, and sub-issues. Either way Copilot reasons over the results directly. What is exclusive is the built-in GitHub MCP server, which needs no token or extension setup; when `gh` is installed, the model mostly uses `gh`, which any agent with a shell can also run. > **Note:** When `gh` is on PATH, Copilot CLI omits these issue tools (`list_issues`, `search_issues`, `issue_read`) and the model is expected to use `gh` instead, > e.g. `gh issue list --state open --label <label>`, `gh issue view <n> --comments`, `gh search issues "<query>"`. To call the MCP tools, > start the session with `--add-github-mcp-tool=<name>` for every GitHub MCP tool you need (it replaces the default set, so include default ones you still use), or `--enable-all-github-mcp-tools`, which also enables write tools. ## When to Use - Triaging a backlog of open issues across one or more repositories - Auto-categorizing issues by content (bug, feature, docs, etc.) - Finding duplicate or related issues before filing new ones - Building one queue from both issues and external-contributor PRs - Linking issues to pull requests for traceability - Generating weekly issue reports or burndown summaries ## Workflow ### 1. List Open Issues with Filters With `gh` on PATH (default), e.g. `gh issue list --repo my-org/my-app --state open --label bug --limit 25`. When the MCP tools are exposed: ```text Tool: github-mcp-server-list_issues owner: "my-org" repo: "my-app" state: "OPEN" labels: ["bug"] orderBy: "UPDATED_AT" direction: "DESC" perPage: 25 ``` ### 2. Search Across Repositories Find issues matching a keyword across your entire org. With `gh` on PATH (default), e.g. `gh search issues "memory leak" --owner my-org --state open`. When the MCP tools are exposed: ```text Tool: github-mcp-server-search_issues query: "memory leak is:open" owner: "my-org" ``` ### 3. Read Issue Details and Comments With `gh` on PATH (default), e.g. `gh issue view 87 --repo my-org/my-app --comments`. When the MCP tools are exposed: ```text Tool: github-mcp-server-issue_read method: "get" owner: "my-org" repo: "my-app" issue_number: 87 Tool: github-mcp-server-issue_read method: "get_comments" owner: "my-org" repo: "my-app" issue_number: 87 ``` ### 4. Check Sub-Issues For tracking epics and parent/child relationships. With `gh` on PATH (default), e.g. `gh api repos/my-org/my-app/issues/42/sub_issues`. When the MCP tools are exposed: ```text Tool: github-mcp-server-issue_read method: "get_sub_issues" owner: "my-org" repo: "my-app" issue_number: 42 ``` ### 5. Get Labels for Classification With `gh` on PATH (default), e.g. `gh issue view 87 --repo my-org/my-app --json labels`. When the MCP tools are exposed: ```text Tool: github-mcp-server-issue_read method: "get_labels" owner: "my-org" repo: "my-app" issue_number: 87 ``` ### 6. Pull External-Contributor PRs into the Queue If the inbox includes PRs as well as issues, filter for outside contributors first and then route those PRs through the same triage state model you use for issues. ```powershell gh pr list --limit 100 --json number,title,authorAssociation,isDraft,url | ConvertFrom-Json | Where-Object { $_.authorAssociation -in @('CONTRIBUTOR', 'FIRST_TIME_CONTRIBUTOR', 'NONE') -and -not $_.isDraft } ``` Treat `OWNER`, `MEMBER`, and `COLLABORATOR` PRs as normal maintainer review unless your project has a stronger triage requirement. If a bare reference such as `#42` could point to either a PR or an issue, check the PR first and fall back to the issue only if no PR exists. When an external PR duplicates existing work, close the loop explicitly: mark it `wontfix`, link the authoritative issue or PR, and avoid creating a second parallel thread for the same request. ## Examples ### Weekly Issue Triage Session > "Show me all open issues in my-org/api-service created in the last 7 days, > grouped by label. For unlabeled issues, suggest a category based on the title > and body." Copilot runs `gh issue list --repo my-org/api-service --search "created:>=<date>"` (when exposed: `list_issues` with a `since` filter), reads each issue's labels, groups them, and uses the issue body to suggest labels for uncategorized ones. ### Duplicate Detection > "Before I file a bug about 'WebSocket disconnects on idle timeout', check > if there's already an issue about this." With `gh` on PATH (default), e.g. `gh search issues "WebSocket disconnect idle timeout" --repo my-org/my-app`. When the MCP tools are exposed: ```text Tool: github-mcp-server-search_issues query: "WebSocket disconnect idle timeout" owner: "my-org" repo: "my-app" ``` Copilot searches, reads the top matches, and tells you whether to file new or comment on an existing issue. ### Sprint Planning Report > "List all issues labeled 'sprint-12' in my-org/frontend, show their status, > assignee, and whether they have linked PRs." Copilot lists the issues, checks each for linked PRs via comments and cross-references, and produces a sprint board summary right in the terminal. ### Shared Issue + PR Intake > "Show me new outside-contributor PRs and new unlabeled issues from the last week, then tell me > which ones need triage first." Copilot lists issues with `gh issue list` (or the MCP issue tools when exposed), pulls in external PRs with `gh pr list`, then routes the highest-signal items into the same triage queue instead of treating PRs as a separate inbox. ### Batch Categorization with SQL Tracking ```sql -- Track triage progress in session database CREATE TABLE IF NOT EXISTS triage_items ( issue_number INTEGER, repo TEXT, title TEXT, suggested_label TEXT, status TEXT DEFAULT 'pending' ); ``` Load issues into the session database, then work through them systematically. ## When the Report Is a Vulnerability A security report must not be triaged in public. If someone opens an issue describing a vulnerability, do not discuss the details in the thread — point the reporter to the repository's private vulnerability reporting (Security → Advisories), then close and lock the public issue. Neither makes its content private: if it contains exploit or other sensitive details, ask a repository admin to delete it — in an organization, an owner may first have to allow issue deletion. For the advisory itself, the agent prepares evidence and fixes; a maintainer accepts, merges, releases, and publishes: | Step | Who | |------|-----| | Verify the claim: who owns the code path, and which release lines are affected | agent, then a maintainer confirms | | Accept the advisory (the reporter sees this) | **human only** | | Fix in the advisory's temporary private fork and review there, not in a public PR | agent may prepare the fix | | Merge to **every** affected release line and release each one | agent lists the lines; a maintainer merges and releases | | Publish the advisory — there is no unpublish | **human only** | After publication, follow-up issues, labels, and board cards fall under the repository's normal triage rules. Do not copy the reproduction or proof of concept onto a project board or a tracking issue: people with project access are not necessarily advisory collaborators. Link to the advisory instead. Before publishing, check the credits and decide on a CVE. For reports that came in through private vulnerability reporting, GitHub adds the reporter as a credited user automatically; for an advisory a maintainer opened, add the credits explicitly. This mirrors the maintainer flow `modelcontextprotocol/inspector` keeps in its repository (documentation under CC-BY-4.0). ## Tips - **Parallel issue reads**: Call `issue_read` (with `gh`: `gh issue view`) on multiple issues simultaneously to build a triage report fast. - **Use `search_issues` (with `gh`: `gh search issues`) for cross-repo queries**: The search syntax supports `org:`, `repo:`, `label:`, `assignee:`, `milestone:`, and date ranges. - **Combine with session SQL**: Store triage results in the session database to track which issues you've reviewed and what actions you took. - **Sub-issues for epic tracking**: Use `get_sub_issues` (with `gh`: `gh api .../issues/<n>/sub_issues`) to check progress on large features broken into subtasks. - **Link to PRs**: After identifying an issue, ask Copilot to search for PRs that reference it — closing the loop between issues and code changes. - **Pair with `triage` for per-item state**: Use [`workflow/triage`](../../workflow/triage/SKILL.md) when one issue or one external PR needs a durable `needs-info` / `ready-for-agent` / `wontfix` decision.
Comments (0)
Sign in to join the conversation.
Reviews (0)
No reviews yet.
No comments yet.