Git PR Reader Skill

Unified interface for GitHub Pull Requests and GitLab Merge Requests — read, review, and post comments.

Capabilities

• Auto-detect Git platform: Automatically identifies GitHub vs GitLab from URL
• Read PR/MR data: Fetch title, description, and file-level diffs with smart filtering
• List changed files: Get file paths, statuses, and line counts with optional glob filtering
• Review comments: Fetch existing review comments/discussions, filter bots, include/exclude resolved
• Post review comments: Post inline comments on specific diff lines with duplicate detection and fallback to PR-level comments
• Reply to comments: Reply to existing review comment threads on GitHub PRs and GitLab MRs
• Extract line numbers: Parse diffs to get added/modified line numbers for accurate comment placement
• Validate comments: Check comment line numbers against actual diff content
• Auto-detect PR/MR: Find the open PR/MR for the current git branch (GitHub via gh CLI, GitLab via API)
• Smart file filtering: Exclude irrelevant files using configurable YAML patterns:
  • Test files (test/, *_test.go, *.spec.ts, etc.)
  • Lock files (package-lock.json, *.lock, Pipfile.lock, etc.)
  • CI/CD configs (.gitlab-ci.yml, .github/workflows/, Dockerfile, etc.)
  • Build artifacts (dist/, build/, target/, *.class, etc.)
  • Generated code (*.pb.go, *.gen.go, etc.)
  • Vendor directories (node_modules/, vendor/, etc.)
  • Images and binaries (*.png, *.jpg, *.svg, etc.)

Usage

Subcommands

read — Read PR/MR data with diffs and file filtering

uv run --script ${CLAUDE_SKILL_DIR}/scripts/git_pr_reader.py read --url "https://github.com/owner/repo/pull/123"
uv run --script ${CLAUDE_SKILL_DIR}/scripts/git_pr_reader.py read --url "https://gitlab.com/group/project/-/merge_requests/456" --format markdown
uv run --script ${CLAUDE_SKILL_DIR}/scripts/git_pr_reader.py read --url "https://github.com/owner/repo/pull/123" --no-filter

info — Get PR/MR information

uv run --script ${CLAUDE_SKILL_DIR}/scripts/git_pr_reader.py info https://github.com/owner/repo/pull/123 --json

files — List changed files

uv run --script ${CLAUDE_SKILL_DIR}/scripts/git_pr_reader.py files https://github.com/owner/repo/pull/123
uv run --script ${CLAUDE_SKILL_DIR}/scripts/git_pr_reader.py files https://github.com/owner/repo/pull/123 --filter "*.adoc" --json

comments — List review comments

uv run --script ${CLAUDE_SKILL_DIR}/scripts/git_pr_reader.py comments https://github.com/owner/repo/pull/123
uv run --script ${CLAUDE_SKILL_DIR}/scripts/git_pr_reader.py comments https://github.com/owner/repo/pull/123 --include-resolved --json

diff — Get unified diff

uv run --script ${CLAUDE_SKILL_DIR}/scripts/git_pr_reader.py diff https://github.com/owner/repo/pull/123

post — Post review comments

uv run --script ${CLAUDE_SKILL_DIR}/scripts/git_pr_reader.py post https://github.com/owner/repo/pull/123 comments.json
uv run --script ${CLAUDE_SKILL_DIR}/scripts/git_pr_reader.py post https://github.com/owner/repo/pull/123 comments.json --review-type technical
uv run --script ${CLAUDE_SKILL_DIR}/scripts/git_pr_reader.py post https://github.com/owner/repo/pull/123 comments.json --review-type style --dry-run

reply — Reply to an existing review comment

# Reply to a GitHub review comment
uv run --script ${CLAUDE_SKILL_DIR}/scripts/git_pr_reader.py reply https://github.com/owner/repo/pull/123 --comment-id 12345 --body "Applied the fix."

# Reply to a GitLab discussion
uv run --script ${CLAUDE_SKILL_DIR}/scripts/git_pr_reader.py reply https://gitlab.com/group/project/-/merge_requests/456 --discussion-id "abc123" --body "Fixed." --signoff "CI bot"

# Dry run
uv run --script ${CLAUDE_SKILL_DIR}/scripts/git_pr_reader.py reply https://github.com/owner/repo/pull/123 --comment-id 12345 --body "Test" --dry-run

extract — Extract line numbers from diff

# Find line number for a pattern
uv run --script ${CLAUDE_SKILL_DIR}/scripts/git_pr_reader.py extract https://github.com/owner/repo/pull/123 path/to/file.adoc "pattern"

# Dump all added/modified lines
uv run --script ${CLAUDE_SKILL_DIR}/scripts/git_pr_reader.py extract --dump https://github.com/owner/repo/pull/123 path/to/file.adoc

# Validate a comments JSON file against the diff
uv run --script ${CLAUDE_SKILL_DIR}/scripts/git_pr_reader.py extract --validate https://github.com/owner/repo/pull/123 comments.json

metadata — Get combined PR/MR metadata

uv run --script ${CLAUDE_SKILL_DIR}/scripts/git_pr_reader.py metadata https://github.com/owner/repo/pull/123
uv run --script ${CLAUDE_SKILL_DIR}/scripts/git_pr_reader.py metadata https://gitlab.com/group/project/-/merge_requests/456
uv run --script ${CLAUDE_SKILL_DIR}/scripts/git_pr_reader.py metadata https://github.com/owner/repo/pull/123 --diff-output /path/to/diff.patch

Returns combined metadata: platform, pr_number, title, description, state, author, base_branch, head_branch, labels, commits, changed_files, and url. With --diff-output, also saves the unified diff to the specified file.

detect — Auto-detect PR/MR for current branch

uv run --script ${CLAUDE_SKILL_DIR}/scripts/git_pr_reader.py detect
uv run --script ${CLAUDE_SKILL_DIR}/scripts/git_pr_reader.py detect --json

Authentication

Set in ~/.env (global) or .env in the project root (local override). See docs-skills README for setup:

GITHUB_TOKEN=your-github-pat    # required scope: "repo" for private, "public_repo" for public
GITLAB_TOKEN=your-gitlab-pat    # required scope: "api"

The script loads .env files automatically — do not prepend source ~/.env to bash commands.

Python Library Usage

from git_pr_reader import GitReviewAPI

api = GitReviewAPI.from_url("https://github.com/owner/repo/pull/123")

# Read PR data with filtering
data = api.get_pr_data()

# Get PR info
info = api.get_pr_info()

# Get changed files
files = api.get_changed_files()

# Get combined metadata (author, labels, commits, files, state)
metadata = api.get_metadata()

# Get review comments
comments = api.get_review_comments()

# Post comments
api.post_comments([{"file": "path/to/file.adoc", "line": 42, "message": "Issue description"}])

# Reply to an existing comment
api.reply_to_comment(comment_id=12345, body="Fixed.")

# Extract line numbers from diff
lines = api.extract_line_numbers("path/to/file.adoc")

# Validate comments against diff
results = api.validate_comments(comments_list)

Comments JSON Format

[
  {"file": "path/to/file.adoc", "line": 42, "message": "Issue: missing title", "severity": "suggestion"},
  {"file": "path/to/other.adoc", "line": 10, "message": "Typo in description", "severity": "suggestion"}
]

Configuration

File filtering patterns are defined in config/git_filters.yaml. You can customize these patterns for your specific needs.

Dependencies

Dependencies (PyGithub, python-gitlab, pyyaml) are declared via PEP 723 inline metadata and installed automatically by uv run --script.

Integration with Other Skills

This skill works well with:

• jira-reader: Combine JIRA issue context with Git code changes
• jira-writer: Generate release notes from JIRA + Git data, then push back to JIRA
• docs-review-style: Post style review findings as inline PR/MR comments
• docs-review-technical: Post technical review findings as inline PR/MR comments

Limitations

• GitHub API rate limits: 60 requests/hour (unauthenticated), 5000/hour (authenticated)
• GitLab API rate limits: 10 requests/second
• Only fetches file-level diffs, not commit-level history
• Requires public PRs/MRs unless authentication tokens are provided