Configure automated Claude Code PR reviews

.claude/review-prompt.md

@@ -0,0 +1,41 @@
+# Claude Code Review Prompt
+
+You are reviewing pull requests for moqueue. Apply the project rules from CLAUDE.md/AGENTS.md (loaded separately) in addition to this guidance.
+
+**Only comment when you have actionable feedback. Never post "looks good", "no issues found", or summary-only comments.**
+
+## What to Evaluate
+
+- **Correctness**: the change should solve the stated problem, handle edge cases, and be free of bugs and regressions.
+- **Code quality, readability, and maintainability**: the change should be clean, easy to follow, and sustainable to maintain and modify.
+- **Conventions**: the change should follow this repo's existing patterns and conventions. It should avoid introducing or perpetuating anti-patterns. Do not re-teach language basics — ground convention feedback in what this codebase actually does.
+- **Performance**: the change should avoid N+1 queries, excess API calls, unnecessary re-renders, unbounded list operations, and avoidable complexity.
+- **Security**: the change should validate and sanitize inputs, enforce authentication and authorization correctly, protect sensitive data, and avoid common web security risks.
+- **Testing**: the change should include coverage for changed behavior, failure modes, and important edge cases. Tests should assert meaningful behavior, not just exercise code.
+- **Design, architecture, and scope**: the change should be well-designed, make worthwhile tradeoffs, fit the rest of the codebase cleanly, and remain reasonably scoped.
+- **Dependencies**: any added library should be necessary and used appropriately.
+
+## Findings
+
+Categorize each finding:
+
+- **Critical**: broken behavior, likely regressions, data loss, security issues, missing authorization, or defects that should block merge
+- **Suggestion**: meaningful improvements to correctness, resilience, performance, maintainability, or test coverage
+- **Nit**: minor style or readability feedback that is clearly optional
+- **Question**: uncertainty about intent, tradeoffs, or missing context
+
+## PR Context
+
+- Link to the project pitch or Jira ticket in the PR body when applicable
+
+## Review Behavior
+
+- Focus on issues that matter. Do not pad the review with trivial comments.
+- Explain what is wrong, why it matters, and suggest a concrete fix when possible. If unsure, say so and phrase it as a question.
+- Flag change hygiene issues when unrelated changes are mixed in or the work should be split into smaller reviewable pieces.
+- Call out genuinely good choices when they are specific and notable.
+- Group related issues into a single review comment
+- Reference specific lines using GitHub line-link format
+- Skip auto-generated files (lockfiles, codegen output, Vite build artifacts)
+- If the entire PR looks good, do not post a comment at all
+- Read all previous review comments before posting. Do not repeat feedback that has already been given — whether by a human reviewer or a prior bot review. If a previous comment flagged an issue and it remains unfixed, you may briefly note it persists but do not re-explain.

.github/workflows/claude-code-review.yml

@@ -0,0 +1,128 @@
+name: Claude Code Review
+
+on:
+  pull_request:
+    types: [opened, synchronize, ready_for_review, reopened]
+  issue_comment:
+    types: [created]
+  pull_request_review_comment:
+    types: [created]
+  pull_request_review:
+    types: [submitted]
+
+permissions:
+  contents: read
+  pull-requests: write
+  issues: read
+  id-token: write
+
+jobs:
+  review:
+    if: |
+      (github.event_name == 'pull_request' && github.event.pull_request.draft == false) ||
+      (github.event_name == 'issue_comment' && github.event.issue.pull_request && contains(github.event.comment.body, '@claude')) ||
+      (github.event_name == 'pull_request_review_comment' && contains(github.event.comment.body, '@claude')) ||
+      (github.event_name == 'pull_request_review' && contains(github.event.review.body, '@claude'))
+    concurrency:
+      group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.event.issue.number || github.run_id }}
+      cancel-in-progress: true
+    runs-on: ubuntu-latest
+    steps:
+      - name: Resolve PR number
+        id: pr
+        run: |
+          if [ "${{ github.event_name }}" = "issue_comment" ]; then
+            echo "number=${{ github.event.issue.number }}" >> "$GITHUB_OUTPUT"
+          else
+            echo "number=${{ github.event.pull_request.number }}" >> "$GITHUB_OUTPUT"
+          fi
+
+      - name: Checkout repository
+        uses: actions/checkout@v4
+        with:
+          ref: refs/pull/${{ steps.pr.outputs.number }}/head
+          fetch-depth: 0
+
+      - name: Fetch PR diff
+        env:
+          GH_TOKEN: ${{ github.token }}
+        run: |
+          gh pr diff "${{ steps.pr.outputs.number }}" > /tmp/pr-diff.patch
+
+      - name: Fetch previous reviews
+        env:
+          GH_TOKEN: ${{ github.token }}
+        run: |
+          PR_NUM="${{ steps.pr.outputs.number }}"
+          REPO="${{ github.repository }}"
+          {
+            echo "=== Review Summaries ==="
+            gh api "repos/${REPO}/pulls/${PR_NUM}/reviews" --paginate \
+              --jq '.[] | select(.state != "PENDING") | "[\(.user.login)] \(.state)\n\(.body)\n---"' 2>/dev/null || true
+
+            echo ""
+            echo "=== Inline Review Comments ==="
+            gh api "repos/${REPO}/pulls/${PR_NUM}/comments" --paginate \
+              --jq '.[] | "[\(.user.login)] \(.path):\(.line // .original_line)\n\(.body)\n---"' 2>/dev/null || true
+
+            echo ""
+            echo "=== PR Conversation Comments ==="
+            gh api "repos/${REPO}/issues/${PR_NUM}/comments" --paginate \
+              --jq '.[] | "[\(.user.login)]\n\(.body)\n---"' 2>/dev/null || true
+          } > /tmp/previous-reviews.txt
+
+      - name: Build review prompt
+        id: prompt
+        run: |
+          DELIMITER="PROMPT_EOF_$(uuidgen)"
+          {
+            echo "REVIEW_PROMPT<<${DELIMITER}"
+            for f in CLAUDE.md AGENTS.md; do
+              if [ -f "$f" ]; then
+                cat "$f"
+                echo ""
+              fi
+            done
+            if [ -f ".claude/review-prompt.md" ]; then
+              cat .claude/review-prompt.md
+            fi
+            echo "${DELIMITER}"
+          } >> "$GITHUB_OUTPUT"
+
+      - name: Resolve claude_args
+        id: args
+        env:
+          MODEL: ${{ vars.CLAUDE_REVIEW_MODEL }}
+        run: |
+          ALLOWED_TOOLS='Read,Glob,Grep,mcp__github_inline_comment__create_inline_comment,Bash(gh pr comment:*),Bash(gh pr view:*)'
+          if [ -n "$MODEL" ]; then
+            echo "value=--model ${MODEL} --max-turns 10 --allowedTools \"${ALLOWED_TOOLS}\"" >> "$GITHUB_OUTPUT"
+          else
+            echo "value=--max-turns 10 --allowedTools \"${ALLOWED_TOOLS}\"" >> "$GITHUB_OUTPUT"
+          fi
+
+      - name: Claude Code Review
+        continue-on-error: true
+        uses: anthropics/claude-code-action@v1
+        with:
+          anthropic_api_key: ${{ secrets.ANTHROPIC_API_KEY }}
+          allowed_bots: "custom-ink-triton"
+          prompt: |
+            You are reviewing PR #${{ steps.pr.outputs.number }}.
+            The full diff is in /tmp/pr-diff.patch — Read that file first.
+            Previous review comments are in /tmp/previous-reviews.txt — Read that file next.
+            Do NOT run gh pr diff or git diff to re-fetch it.
+            Use Read on source files if you need more context beyond the diff.
+
+            IMPORTANT: Do NOT repeat or reiterate feedback that has already been given
+            in previous reviews. If a prior reviewer (human or bot) already flagged an
+            issue and the author has not addressed it in the current diff, you may
+            reference it briefly but do not re-explain. Focus your review on NEW issues
+            not covered by prior feedback.
+
+            Post inline review comments and stop.
+
+            Review this pull request using the following guidelines:
+
+            ${{ steps.prompt.outputs.REVIEW_PROMPT }}
+          claude_args: "${{ steps.args.outputs.value }}"

AGENTS.md

@@ -0,0 +1,26 @@
+# Moqueue
+
+Ruby gem for mocking the AMQP library in tests. Allows testing AMQP-based code without running a broker by providing mock replacements for queues, exchanges, and channels.
+
+## Tech Stack
+
+- **Language:** Ruby (see `moqueue.gemspec` for dependencies)
+- **Type:** Gem (consumed via Bundler)
+- **Key dependency:** `amqp` >= 0.9.0
+
+## Local Development & Testing
+
+Run `bundle install` and `bundle exec rake spec`. Test framework: RSpec.
+
+## Key Directories
+
+| Path | Purpose |
+|------|---------|
+| `lib/moqueue/` | Mock implementations (MockQueue, MockExchange, MockChannel) |
+| `spec/` | RSpec suite |
+
+## Architecture Notes
+
+- Provides `overload_amqp` to globally replace AMQP classes with mocks
+- Captures published messages in a `routed_messages` array for assertions
+- Used by other Custom Ink services (e.g. journal-service) for testing AMQP workflows

CLAUDE.md

@@ -0,0 +1,3 @@
+This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.
+
+@AGENTS.md