check-pr

name	check-pr
description	Dùng khi user runs `/check-pr` hoặc nói 'check PR', 'verify CI', 'kiểm tra PR', 'ready to merge?'. Monitor CI via scripts (`scripts/check-ci.sh`), run local E2E + unit tests if Docker / JAVA_HOME available, output verdict (✅ ready / ❌ fix needed / ⚠️ conditional).
user-invocable	true
argument-hint	[PR-number|branch-name]

/check-pr — Monitor CI & Verify PR Quality

Usage

/check-pr 207              # Check PR #207
/check-pr wave/3           # Check branch wave/3
/check-pr                  # Check current branch

---

Rules

1. MCP-first — dùng GitHub MCP nếu connected (claude mcp list), gh CLI làm fallback. Chi tiết: .claude/rules/mcp-first-with-fallback.md
2. LUÔN dùng scripts — không chạy gh run list trực tiếp nếu có script
3. E2E verify ở local — Docker phải up, chạy ./scripts/test-api-e2e.sh
4. CI monitor bằng script hoặc MCP — GitHub MCP list_workflow_runs hoặc bash scripts/check-ci.sh [branch]
5. Cleanup bằng script — bash scripts/cleanup-ci-runs.sh

---

Step 1: Identify Target

# Nếu PR number:
PR_NUMBER=$ARGUMENTS
BRANCH=$(gh pr view $PR_NUMBER --json headRefName --jq '.headRefName')

# Nếu branch name:
BRANCH=$ARGUMENTS

# Nếu không có argument:
BRANCH=$(git branch --show-current)

---

Step 2: Monitor CI (bằng script)

# LUÔN dùng script
bash scripts/check-ci.sh $BRANCH

# Nếu timeout → tăng thời gian:
bash scripts/check-ci.sh $BRANCH 15  # 15 phút timeout

Script location: scripts/check-ci.sh Behavior: Poll mỗi 15s, hiển thị status, exit 0 nếu all pass, exit 1 nếu fail.

Lưu ý: Script check TẤT CẢ runs trên branch (kể cả cũ). Nếu có failure cũ → check timestamp để phân biệt.

---

Step 3: Analyze Failures (nếu CI fail)

# Tìm failed run ID
FAILED_RUN=$(gh run list --branch $BRANCH --limit 5 \
  --json databaseId,workflowName,conclusion \
  --jq '.[] | select(.conclusion=="failure") | .databaseId' | head -1)

# Xem log lỗi
gh run view $FAILED_RUN --log-failed 2>/dev/null | tail -30

# Tìm root cause
gh run view $FAILED_RUN --log-failed 2>/dev/null \
  | grep -E "ERROR|FAIL|Compilation|cannot find|Tests run.*Failures: [1-9]|Tests run.*Errors: [1-9]" \
  | head -10

---

Step 4: Local Verification (TRƯỚC khi merge)

4a. E2E Tests (nếu Docker running)

# Check Docker status trước
cd kitehub && ./scripts/status.sh --health 2>/dev/null

# Nếu Docker up:
./scripts/wait-for-healthy.sh
./scripts/test-api-e2e.sh

# Nếu Docker down → ghi nhận:
# "⚠️ Docker not running — E2E skipped, rely on CI"

4b. Unit Tests (nếu JAVA_HOME available)

# KiteHub
cd kitehub && JAVA_HOME=/home/vkiet/jdk/jdk-21 ./mvnw clean test -pl kitehub-subscription -am -q

# KiteClass
cd kiteclass/kiteclass-core && JAVA_HOME=/home/vkiet/jdk/jdk-21 ./mvnw clean test -q

4c. Frontend Build

cd kitehub/kitehub-frontend && pnpm build
cd kiteclass/kiteclass-frontend && pnpm build

Note: Nếu không thể chạy local → ghi nhận và rely on CI. KHÔNG block merge vì thiếu local test — CI là minimum.

---

Step 5: Output Report

## PR Check Report: #[number] ([branch])

### CI Status
- [ ] All workflows: ✅/❌ (via `scripts/check-ci.sh`)
- Failed: [list if any]

### Local Verification
- [ ] E2E tests: ✅/❌/⏭️ skipped (Docker: up/down)
- [ ] Unit tests: ✅/❌/⏭️ skipped (JAVA_HOME: set/unset)
- [ ] Frontend build: ✅/❌/⏭️ skipped

### Issues Found
[Any failures with root cause]

### Verdict
✅ Ready to merge
❌ Fix needed: [description]
⚠️ Conditional: CI pass but local not verified

---

Step 6: Merge (CHỈ khi verdict = ✅)

KHÔNG merge nếu verdict = ❌ hoặc ⚠️ chưa resolve.

6a. Phân biệt merge target

User nói "merge" → merge agent PRs vào WAVE BRANCH
User nói "merge wave → main" hoặc "merge vào main" → merge wave → main
KHÔNG BAO GIỜ tự suy diễn merge vào main

6b. Xác nhận user approve

LUÔN hỏi trước khi merge:

• PR → wave: "Merge 4 PRs vào wave/X?"
• Wave → main: "Wave check pass. Merge wave/X → main?"
• KHÔNG tự merge bất kỳ thứ gì mà không có confirm rõ ràng.

6c. Merge command

# PR agent → wave branch (user đã confirm):
gh pr merge [number] --squash --delete-branch

# Wave → main (user đã confirm RÕ RÀNG "merge vào main"):
gh pr merge [number] --squash --delete-branch
git checkout main && git pull

# Hotfix → main (user đã confirm):
gh pr merge [number] --squash --delete-branch

6d. VIOLATION LOG (Wave 4, 2026-03-24)

User nói "merge" → agent merge cả wave/4 → main mà không hỏi. Đúng ra: chỉ merge 4 PRs vào wave/4, rồi HỎI user trước khi wave/4 → main.

6c. Verify sau merge

# Sync local
git checkout [target-branch] && git pull

# Verify CI triggered trên target
bash scripts/check-ci.sh [target-branch] 3

---

Step 7: Cleanup (sau merge)

# Clean stale branches
git remote prune origin

# Clean worktrees
rm -rf .claude/worktrees/agent-* 2>/dev/null
git worktree prune

# Check remaining
echo "Stale branches:" && git branch -r | grep -v "main\|HEAD\|wave/" | wc -l
echo "Open PRs:" && gh pr list --state open --json number --jq 'length'

---

Integration với Wave Process

Per PR:  /check-pr [PR-number]          ← Sau agent tạo PR
Wave:    /check-pr wave/X               ← Sau merge tất cả PRs vào wave
Main:    /check-pr main                 ← Sau merge wave → main
Full:    /wave-completion-check [X]     ← 6-level check

---

Gotchas

• gh pr checks --json state returns null — use scripts/check-ci.sh or --json status instead; raw state field breaks polling loops silently (memory feedback_monitor_gh_pr_checks_no_state_field.md)
• Step 2 script picks up STALE failures — scripts/check-ci.sh shows ALL runs on the branch including ancient ones; cross-check timestamps before declaring "CI red" (per Step 2 note in body)
• Stacked PR --delete-branch trap — merging a stacked PR with --delete-branch auto-closes child PRs and they cannot be reopened (memory feedback_stacked_pr_delete_branch.md); for stacks merge parent without --delete-branch first, then re-target child
• "merge" never auto-promotes wave → main — Step 6 is explicit but easy to skip: user saying "merge" means agent PRs → wave branch ONLY; wave → main requires separate explicit confirmation (Wave 4 violation 2026-03-24 logged in §6d)
• Local skip ≠ block — if Docker down or JAVA_HOME unset, mark E2E/unit ⏭️ skipped and rely on CI; verdict can still be ✅ Ready (per Step 4 note); do NOT downgrade to ❌ just because local couldn't run