CodeRabbit Intake

Minimal pipeline for CodeRabbit-created GitHub issues. These issues are already well-structured with clear context, file locations, and actionable tasks — no heavy research needed.

When to Use

• User says "do coderabbit issues", "pick up a coderabbit ticket", "work on coderabbit issues"
• During issue-triage when author.login === "app/coderabbitai"
• Direct invocation with a coderabbit issue URL

Configuration

# Target repository (checked in order):
# 1. $CR_TARGET_REPO env var (e.g., "Comfy-Org/ComfyUI_frontend")
# 2. Git remote origin if in a repo
# 3. Prompt user

Workflow

Step 1: Determine Target Repository

# Check env var first
if [ -n "$CR_TARGET_REPO" ]; then
  REPO="$CR_TARGET_REPO"
# Try git remote
elif git remote get-url origin 2>/dev/null; then
  REPO=$(git remote get-url origin | sed 's/.*github.com[:/]\(.*\)\.git/\1/' | sed 's/.*github.com[:/]\(.*\)/\1/')
else
  # Prompt user
fi

Step 2: Find Newest Unclaimed Issue

# Find newest open coderabbit issue without "in progress" label
# Note: --author flag doesn't work reliably for bot accounts, so filter with jq
gh issue list --repo "$REPO" \
  --state open \
  --json number,title,url,labels,createdAt,author \
  --limit 30 | jq '[.[] | select(.author.login == "app/coderabbitai") | select(.labels | map(.name) | index("in progress") | not)] | sort_by(.createdAt) | reverse | .[0]'

If no issues found:

✅ No unclaimed CodeRabbit issues in $REPO.

Step 3: Fetch Issue Details

gh issue view $NUMBER --repo "$REPO" --json title,body,labels,url,createdAt

Step 4: Assess Complexity

Parse the issue body and determine complexity:

Complex indicators (suggest full pipeline):

• 5+ files mentioned across different directories
• Keywords: "centralize", "audit", "throughout codebase", "refactor all", "create new system", "implement store", "add rollback"
• Vague scope without specific file references
• Multiple acceptance criteria
• New architecture/patterns required

Simple indicators (terminal branch OK):

• 1-3 specific files listed
• Keywords: "fix type", "rename", "update", "align with", "follow pattern", "localize", "replace"
• Clear single transformation (e.g., "replace X with Y")
• Single PR/comment reference with clear instruction

Score: Count complex vs simple signals.

Step 5: Present Choice

## CodeRabbit Issue Found

**Issue:** #$NUMBER - $TITLE
**Created:** $DATE
**URL:** $URL

### Complexity Assessment

$ASSESSMENT_REASONING

---

**Choose approach:**

**(a) Full Pipeline** — Use `ticket-intake` for research, planning, and structured implementation
→ Load: `ticket-intake` with issue URL

**(b) Terminal Branch** — Implement directly, quality-checks, PR
→ Continue below

Wait for user input.

Step 6: Claim Issue (Terminal Branch)

If user chooses (b):

# Get current GitHub username
USERNAME=$(gh api user --jq '.login')

# Add "in progress" label
gh issue edit $NUMBER --repo "$REPO" --add-label "in progress"

# Assign to self
gh issue edit $NUMBER --repo "$REPO" --add-assignee "$USERNAME"

Step 7: Output Implementation Context

Parse and structure the issue body:

## Implementation Context

**Issue:** #$NUMBER - $TITLE
**Claimed by:** @$USERNAME
**Status:** in progress

### Task

$EXTRACTED_TASK_DESCRIPTION

### Affected Files

- `$FILE1` — $WHAT_TO_DO
- `$FILE2` — $WHAT_TO_DO

### References

- **Originating PR:** $PR_LINK
- **Discussion:** $COMMENT_LINK
- **Requested by:** @$REQUESTER

### Code Context

$ANY_CODE_SNIPPETS_FROM_ISSUE

---

## Next Steps

1. **Implement** the changes described above
2. **Run quality checks:**

Load the quality-checks skill

3. **Create PR:**
```bash
gh pr create --repo $REPO --fill --body "Fixes #$NUMBER

## Summary
[Brief description]

## Changes
- [Change 1]
- [Change 2]"

### Step 8: Hand Off to Full Pipeline (if chosen)

If user chooses (a):

```markdown
Handing off to full pipeline...

Load the `ticket-intake` skill with:
$ISSUE_URL

Parsing CodeRabbit Issue Body

CodeRabbit issues follow a consistent structure:

## Context

[Link to PR and comment]

## Issue / Problem / Goal

[Description]

## Affected Files / File Location / Locations

- `path/to/file.ts`

## References / Related

- PR: #1234
- Discussion: [link]
- Requested by: @user

Extract:

• Task: From "Issue", "Problem", "Goal", "Task", "Requested Action" sections
• Files: From "Affected Files", "File Location", "Locations" sections
• PR Reference: URLs matching github.com/.*/pull/\d+
• Comment Reference: URLs matching #discussion_r\d+
• Requester: @username after "Requested by"

Error Handling

No "in progress" Label

If the label doesn't exist in the repo:

⚠️ Label "in progress" not found in $REPO.

Creating it now...
gh label create "in progress" --repo "$REPO" --color "FFA500" --description "Issue is being worked on"

Issue Already Claimed

If the issue already has "in progress" label or assignee, skip and find next:

Issue #$NUMBER already claimed. Finding next...

Auth Issues

❌ GitHub CLI not authenticated or lacks permissions.
Run: gh auth login

Notes

• This is a terminal branch — no runs/ directory or persistent state
• Reuses existing quality-checks skill for verification
• PR creation uses standard gh pr create with issue link
• Complexity assessment is heuristic — human makes final call

Integration with Issue Triage

During issue-triage sessions, CodeRabbit issues should be skipped (not individually triaged):

• Author is coderabbitai[bot] or app/coderabbitai
• These are auto-generated refactoring suggestions
• Handle them in batches using this skill, not during triage

Batch Handling Suggestions

When many CodeRabbit issues accumulate (20+), suggest to maintainers:

1. Batch-close low-priority: Close refactoring issues that aren't urgent
2. Tech debt document: Convert to a tracking document instead of individual issues
3. Good first issue: Tag actionable ones with good first issue for contributors
4. Triage session: Run dedicated CodeRabbit intake session to clear backlog

Finding All CodeRabbit Issues

# List all open CodeRabbit issues
gh issue list --repo "$REPO" --state open --limit 100 \
  --json number,title,createdAt,labels \
  | jq '[.[] | select(.author.login == "coderabbitai" or .author.login == "app/coderabbitai")]'