pr-workflow

name	pr-workflow
description	ALWAYS USE THIS SKILL when: creating PRs, checking PR status, reviewing PRs, merging PRs, checking CI status, fixing lint/clippy errors, running cargo fmt, or any gh pr command. Invoke with /pr-workflow
allowed-tools	Bash, Read, Grep, Glob
user-invocable	true

Pull Request Workflow for Rust Projects

Quick Reference

Task	Command
CI status (non-blocking)	gh pr view <N> --json statusCheckRollup --jq '.statusCheckRollup[] | "\(.name): \(.conclusion // "pending")"'
CI status (blocking)	gh pr checks <N>
Read PR comments (Claude)	gh pr view <N> --json comments --jq '.comments[] | "---\n" + .body'
Read inline comments (Codex)	gh api repos/$REPO/pulls/<N>/comments --jq '.[] | select(.user.login | test("codex")) | "---\nfile: \(.path):\(.line // .original_line)\n\(.body)"'
Read all inline comments	gh api repos/$REPO/pulls/<N>/comments --jq '.[] | "---\nfile: \(.path):\(.line // .original_line)\n\(.body)"'
Create PR	git push -u origin <branch> && gh pr create --fill
Merge standalone PR	gh pr merge <N> --merge --delete-branch
Merge stacked PR (base)	gh pr merge <N> --merge (NO --delete-branch!)
List my PRs	gh pr list --author @me

Setup — run this once per session to avoid placeholder issues:

REPO=$(gh repo view --json nameWithOwner --jq '.nameWithOwner')

---

CI Failed — What Do I Do?

This is the most common use of this skill. Follow this decision tree:

Step 1: Get the run ID and identify failures

REPO=$(gh repo view --json nameWithOwner --jq '.nameWithOwner')

# Get latest run ID for current branch
RUN_ID=$(gh run list --branch "$(git branch --show-current)" --limit 1 --json databaseId --jq '.[0].databaseId')

# Show all failed jobs and which step failed
gh api "repos/$REPO/actions/runs/$RUN_ID/jobs" \
  --jq '.jobs[] | select(.conclusion == "failure") | "\(.id) \(.name): \([.steps[] | select(.conclusion == "failure") | .name] | join(", "))"'

Step 2: Is it your fault?

# Check if main is green
gh run list --branch main --limit 3 --json conclusion --jq '.[].conclusion'

• Main green, PR red → Your PR caused it. Fix it.
• Main also red → Pre-existing. Note in PR, but still investigate.

Step 3: Get logs for the failed job

gh run view --job --log blocks until the ENTIRE run finishes. Use the Jobs API instead:

JOB_ID=<from step 1>

# Get full log (works immediately for completed jobs)
gh api "repos/$REPO/actions/jobs/$JOB_ID/logs" 2>&1 | tail -100

# Search for errors
gh api "repos/$REPO/actions/jobs/$JOB_ID/logs" 2>&1 | grep -E "error|FAIL|panicked" | head -20

# Get context around failures
gh api "repos/$REPO/actions/jobs/$JOB_ID/logs" 2>&1 | grep -B5 "Error" | head -30

Step 4: Fix and push

# Fix the code, then:
cargo fmt
cargo clippy --all-targets -- -D warnings
git add <files> && git commit -m "fix: ..."
git push

Step 5: Report with evidence, not speculation

NEVER say "likely", "probably", or "may be caused by". Always find the actual root cause.

• Read the diff (git diff main..HEAD) to understand what changed
• Match log errors to specific lines in the diff
• If a CI step name is misleading (e.g., "container-test" runs host builds first), read the full log to find where the error actually occurs
• State what the error IS, not what it "might be"

Common failure patterns

• Short failure (<30s) in build step → Compilation error. Get logs immediately.
• "container-test" failure → Check whether error is in host build (runs first) or container build/test (runs after). Read the log.
• Test timeout (600s+) → Test hang. Check for pipe deadlock or missing stdin EOF.

---

Before Merging: Read ALL PR Comments

MANDATORY before any merge, push, or PR update. Run ALL THREE commands:

REPO=$(gh repo view --json nameWithOwner --jq '.nameWithOwner')

# PR-level comments (Claude review posts here)
gh pr view <N> --json comments --jq '.comments[] | "---\n" + .body'

# Inline code review comments (Codex review posts here)
gh api "repos/$REPO/pulls/<N>/comments" --jq '.[] | "---\nfile: \(.path):\(.line // .original_line)\n\(.body)"'

# PR reviews with body text (Codex review summary lives here)
gh api "repos/$REPO/pulls/<N>/reviews" --jq '.[] | select(.body != "") | "---\nauthor: \(.user.login)\n\(.body)"'

Two Automated Reviewers

This repo has two automated code reviewers. Read findings from BOTH before merging.

Claude Review (claude-claude bot):

• Posts as a PR-level comment (visible via gh pr view --json comments)
• Runs as a GitHub Actions workflow on every push
• Severity levels: LOW, MEDIUM, HIGH

Codex Review (chatgpt-codex-connector[bot]):

• Posts a PR review with inline comments on specific lines
• Review summary: gh api repos/$REPO/pulls/<N>/reviews (look for chatgpt-codex-connector)
• Inline suggestions: gh api repos/$REPO/pulls/<N>/comments (filter by codex author)
• Priority badges: P0 (critical), P1 (important), P2 (suggestion)

Read Codex inline comments specifically:

# Codex inline review comments only
gh api "repos/$REPO/pulls/<N>/comments" \
  --jq '.[] | select(.user.login | test("codex")) | "---\nfile: \(.path):\(.line // .original_line)\n\(.body)"'

Check for auto-fix PRs

gh pr list --search "base:$(git branch --show-current)"

Decision tree for auto-fix PRs:

Auto-fix PR exists?
 → Run the failing test locally WITHOUT the auto-fix
   PASSES → Close the auto-fix (your branch already fixes it)
            gh pr close <fix-pr> --comment "Not needed - <explain>"
   FAILS  → Read the auto-fix diff
            Adds skip/ignore/early-return? → Close it, find real fix
            Fixes a real gap?              → Cherry-pick, push, close auto-fix

Fix ALL review findings (including low severity) before merging.

---

Merging a PR

Standalone PR (base is main)

gh pr merge <N> --merge --delete-branch

Stacked PR (base is another branch)

NEVER use --delete-branch on the base PR of a stack. It deletes the branch before GitHub retargets dependent PRs, causing them to auto-close.

# Step 1: Merge WITHOUT --delete-branch
gh pr merge <base-pr> --merge

# Step 2: Retarget dependent PR
gh pr edit <dependent-pr> --base main

# Step 3: Verify retarget
gh pr view <dependent-pr> --json baseRefName
# Must show: {"baseRefName":"main"}

# Step 4: NOW delete the old branch
git push origin --delete <base-branch-name>

Recovery if you accidentally used --delete-branch:

git push origin <dependent-branch>  # Re-push (still exists locally)
gh pr create --base main --title "..." --body "..."

---

Creating a Pull Request

1. Run pre-push checks

cargo fmt --check    # Fix: cargo fmt
cargo clippy --all-targets -- -D warnings
make test-root FILTER=sanity  # Or tests relevant to your changes

2. Review and push

git diff --stat                       # What files changed
git log main..HEAD --oneline          # All commits in this branch
git push -u origin <branch-name>
gh pr create --fill                   # For single-commit PRs

For multi-commit PRs, write a structured description:

gh pr create --title "..." --body "$(cat <<'EOF'
## Summary
- Change 1
- Change 2

## Test Results
$ make test-root FILTER=sanity
<actual output>
EOF
)"

3. Verify description matches ALL commits

Do this on creation AND after pushing new commits:

# For PRs targeting main:
git log --oneline origin/main..HEAD

# For stacked PRs (targeting another branch):
BASE=$(gh pr view --json baseRefName --jq '.baseRefName')
git log --oneline "origin/$BASE..HEAD"

# Update if needed:
gh pr edit <N> --body "$(cat <<'EOF'
...updated description covering ALL commits...
EOF
)"

Anti-patterns: Description only covers first/last commit, mentions reverted changes, or (for stacked PRs) includes parent branch commits.

4. Wait for CI

gh pr checks <N>
# All checks must show "pass" before proceeding

---

Stacked PRs (Branch of Branch)

git checkout pr1-branch
git checkout -b pr2-branch
# ... make changes ...
git push -u origin pr2-branch
gh pr create --base pr1-branch

Rebasing after base PR merges

Never use plain git rebase origin/main — it replays already-merged commits, causing false conflicts in files you didn't touch.

# Step 1: Find which commits are yours vs already-merged
git log --oneline origin/main..your-branch

# Step 2: Find the old base tip (last already-merged commit)
for commit in $(git log --format="%h" origin/main..your-branch); do
  subject=$(git log --format="%s" -1 $commit)
  match=$(git log --oneline origin/main --grep="$subject" | head -1)
  echo "$commit: $subject → ${match:-UNIQUE}"
done

# Step 3: Rebase only YOUR commits onto main
git rebase --onto origin/main <old-base-tip> your-branch

# Step 4: Force push
git push origin your-branch --force-with-lease

For deeper stacks, rebase bottom-up:

git rebase --onto pr2-branch <pr2-old-tip> pr3-branch
git push origin pr3-branch --force-with-lease

---

CI Not Starting?

Most common cause: PR branch not based on main.

gh pr view <N> --json baseRefName           # Check target branch
git fetch origin && git rebase origin/main  # Rebase if stale
git push --force-with-lease
gh pr edit <N> --base main                  # Fix wrong target

---

SSH into CI Runners

# Find runners
gh api repos/ejc3/fcvm/actions/runners \
  --jq '.runners[] | "\(.name) busy=\(.busy) labels=\(.labels | map(.name) | join(","))"'

# Get runner IPs (change architecture filter as needed: x86_64 or arm64)
aws ec2 describe-instances \
  --filters "Name=tag:Name,Values=*runner*" "Name=architecture,Values=arm64" \
  --query 'Reservations[*].Instances[*].[InstanceId,PublicIpAddress,State.Name]' --output text

# SSH in
ssh -i ~/.ssh/runner_key -o StrictHostKeyChecking=no ubuntu@<ip>
cd ~/fcvm && git fetch && git checkout <branch>
make build && sudo make test-root FILTER=<test> STREAM=1 2>&1 | tee /tmp/test.log

Don't interfere with busy runners (check busy= status first).

---

Downloading Test Log Artifacts

CI uploads test logs as artifacts (14 day retention). Use when you need full debug logs beyond what the job log provides.

REPO=$(gh repo view --json nameWithOwner --jq '.nameWithOwner')
RUN_ID=<run_id>

# List artifacts
gh api "repos/$REPO/actions/runs/$RUN_ID/artifacts" \
  --jq '.artifacts[] | "\(.id) \(.name) \(.size_in_bytes)"'

# Download a specific artifact
gh api "repos/$REPO/actions/runs/$RUN_ID/artifacts/<artifact_id>/zip" > /tmp/artifact.zip
unzip -o /tmp/artifact.zip -d /tmp/ci-logs/

# Download ALL test log artifacts
for artifact in $(gh api "repos/$REPO/actions/runs/$RUN_ID/artifacts" \
  --jq '.artifacts[] | select(.name | startswith("test-logs")) | "\(.id):\(.name)"'); do
  ID=${artifact%%:*}; NAME=${artifact#*:}
  echo "Downloading $NAME..."
  gh api "repos/$REPO/actions/runs/$RUN_ID/artifacts/$ID/zip" > "/tmp/$NAME.zip"
  mkdir -p "/tmp/ci-logs/$NAME"
  unzip -o "/tmp/$NAME.zip" -d "/tmp/ci-logs/$NAME"
done

What to look for: Full error messages (CI truncates stderr), timestamps (compare failing vs passing for contention), boot sequence (fc-agent startup stalls), dmesg artifacts.