fetch-milestone-prs

This skill retrieves all merged pull requests associated with a GitHub milestone and saves each PR's metadata as a separate JSON file for downstream processing (e.g., release notes generation, changelog creation).

When to Use This Skill

User asks to fetch or retrieve PRs for a milestone
User wants to gather PR data before generating release notes
User needs milestone PR metadata for analysis or categorization
User mentions a milestone name and wants to see what PRs were included
As a prerequisite step before invoking the release-notes prompt

Prerequisites

The gh CLI must be installed and authenticated (gh auth login)
Python 3.8+ must be available

Instructions

Identify the milestone name from the user's request. This is typically a version string like 7.0.0-preview4, 6.1.5, etc.

Identify the repository (optional). Defaults to dotnet/SqlClient. If the user specifies a different repo, pass it via --repo.

Run the fetch script located in this skill's directory:

python .github/skills/fetch-milestone-prs/fetch-milestone-prs.py <milestone> [--repo OWNER/REPO] [--output-dir DIR]

Examples:

# Default: outputs to .milestone-prs/<milestone>/
python .github/skills/fetch-milestone-prs/fetch-milestone-prs.py 7.0.0-preview4

# Custom output directory
python .github/skills/fetch-milestone-prs/fetch-milestone-prs.py 7.0.0-preview4 --output-dir ./my-prs

# Different repo
python .github/skills/fetch-milestone-prs/fetch-milestone-prs.py 7.0.0-preview4 --repo dotnet/efcore

Verify the output by checking the generated _index.json file in the output directory. It contains a summary of all fetched PRs.
Report results to the user:
- Total merged PRs found
- Total skipped (closed but not merged)
- Output directory path
- Mention that each PR is in a separate <number>.json file and the index is at _index.json

Output Format

Directory structure

.milestone-prs/<milestone>/
├── _index.json          # Summary index of all PRs
├── 1234.json            # Individual PR metadata
├── 1235.json
└── ...

Individual PR file (`<number>.json`)

Each file contains:

{
  "number": 1234,
  "title": "PR title",
  "author": "github-username",
  "author_association": "MEMBER",
  "labels": ["label1", "label2"],
  "assignees": ["user1"],
  "state": "closed",
  "merged": true,
  "merged_at": "2026-01-15T12:00:00Z",
  "merge_commit_sha": "abc123...",
  "created_at": "2026-01-10T10:00:00Z",
  "closed_at": "2026-01-15T12:00:00Z",
  "html_url": "https://github.com/dotnet/SqlClient/pull/1234",
  "body": "Full PR description markdown...",
  "comments_count": 5,
  "is_merged_pr": true,
  "has_public_api_label": false,
  "has_engineering_label": false,
  "has_test_label": true
}

Index file (`_index.json`)

{
  "milestone": "7.0.0-preview4",
  "repo": "dotnet/SqlClient",
  "total_closed": 76,
  "total_merged": 74,
  "total_skipped": 2,
  "prs": [
    {
      "number": 1234,
      "title": "PR title",
      "author": "github-username",
      "labels": ["label1"],
      "merged_at": "2026-01-15T12:00:00Z",
      "has_public_api_label": false,
      "has_engineering_label": false,
      "has_test_label": true
    }
  ]
}

Derived Fields

The script computes these boolean flags for easy categorization:

Field	Logic
`has_public_api_label`	Any label contains "Public API :new:"
`has_engineering_label`	Any label contains "Engineering"
`has_test_label`	Any label contains "Test"

Error Handling

If the milestone is not found, the script prints available milestones and exits. Ask the user to verify the milestone name.
If gh is not authenticated, the script will fail with an auth error. Instruct the user to run gh auth login.
If a specific PR detail fetch fails, the script exits. This is rare but can happen with API rate limits — suggest waiting and retrying.
On Windows, if encoding errors occur, the script uses errors="replace" to handle non-UTF-8 characters gracefully.

Using Output with Other Workflows

The output files are designed to be consumed by:

The release-notes prompt — tag the output directory (e.g., @.milestone-prs/7.0.0-preview4/*) when invoking it
Manual analysis — read _index.json for a quick overview, individual files for PR details
Any script or agent that needs structured PR metadata