PR Comments — Fetch, Display, and Fix

Fetch and display comments from the current branch's pull request. Optional argument: $ARGUMENTS can be a PR number, fix to automatically address review comments, or fix 12345 to fix comments on a specific PR.

Step 1: Determine the PR

bash
# Extract PR number from arguments (handles "fix", "fix 12345", "12345", or empty)
PR_NUMBER=$(echo "$ARGUMENTS" | grep -o '[0-9]\+' | head -n 1)
if [[ -z "$PR_NUMBER" ]]; then
  PR_NUMBER=$(gh pr view --json number -q .number)
fi

If no PR is found, stop and tell the user: "No open PR found for the current branch. Usage: /pr-comments [PR_NUMBER]"

Step 2: Fetch All Comments

Run these commands to gather all comment types:

bash
REPO=$(gh repo view --json nameWithOwner -q .nameWithOwner)

# PR-level comments (issue comments)
gh api /repos/${REPO}/issues/${PR_NUMBER}/comments

# Code review comments (inline on diff)
gh api /repos/${REPO}/pulls/${PR_NUMBER}/comments

Step 3: Display Comments

Format and display all comments in a readable way:

PR-level Comments

- @author (date):
  > comment body

Code Review Comments (inline)

- @author file.go#line:
  ```diff
  [diff_hunk]
  ```
  > comment body

Rules:

• Filter out bot noise (labeler messages, generic bot summaries) — but keep actionable bot comments (code suggestions, warnings)
• Preserve threading/nesting of replies
• Show file and line number context for code review comments
• Group comments by file when there are multiple

Step 4: Analyze and Fix (when $ARGUMENTS contains "fix")

If the user invoked /pr-comments fix or asks to fix the comments:

1. Categorize each review comment:

• Actionable code change: Reviewer suggests a specific code change (e.g., add a parameter, rename, fix logic)
• Question/discussion: Reviewer asks a question or raises a discussion point — skip these, surface to user
• Bot suggestion: Automated tool suggests a change — evaluate if valid before applying

1. For each actionable comment:

• Read the file referenced in the comment
• Understand the reviewer's suggestion in context of the full file
• Apply the fix using Edit tool
• Track which comments were addressed

1. After all fixes:

• Run the appropriate validation for affected services (see service map below)
• Commit the changes with message: fix(scope): address PR review feedback (derive scope from the modified file paths using the scope map in the create-pr skill — e.g., api-server/services/ → NB-xxx, app/ → ui)
• Push to the branch

1. Post a reply on each addressed comment:

bash # For inline review comments (code review comments): gh api /repos/${REPO}/pulls/${PR_NUMBER}/comments/${COMMENT_ID}/replies -f body="Fixed in $(git rev-parse --short HEAD)"

# For PR-level comments (issue comments): gh api /repos/${REPO}/issues/${PR_NUMBER}/comments -f body="Addressed in $(git rev-parse --short HEAD)"

1. Report to user:

Addressed N review comments:

• file.go#L42: <short description of fix>
• file.go#L88: <short description of fix>

Skipped M comments (questions/discussion):

• @reviewer: "question text..." — needs your input

Pushed commit: <sha>

Service Validation Map

| Path prefix | Validation command | |---|---| | api-server/services/ | cd api-server/services && make validate | | ticket-server/ | cd ticket-server && make validate | | collector-server/cloud-collector/ | cd collector-server/cloud-collector && make validate | | collector-server/k8s-collector/relay-server/ | cd collector-server/k8s-collector/relay-server && make validate | | collector-server/k8s-collector/app/ | cd collector-server/k8s-collector/app && make lint && make test | | llm/code-analysis/ | cd llm/code-analysis && make check | | llm/llm-server/ | cd llm/llm-server && make validate | | llm/rag-server/ | cd llm/rag-server && make lint && make test | | ml-k8s-server/ | cd ml-k8s-server && make lint && make test | | auto-pilot/ | cd auto-pilot && poetry run black --check . && poetry run flake8 . | | notifications-server/ | cd notifications-server && poetry run black --check . && poetry run flake8 . | | app/ | cd app && npm run lint2 |