AgentStack
Browse Sign in
Browse Why AgentStack Sell Docs
Sign in
SKILL verified Apache-2.0 Self-run

Coderabbit Review

skill-opendatahub-io-ai-helpers-coderabbit-review · by opendatahub-io

Use when you need to evaluate CodeRabbit PR comments and fix or reply

No reviews yet
0 installs
34 views
0.0% view→install

Install

$ agentstack add skill-opendatahub-io-ai-helpers-coderabbit-review

✓ scanned · ✓ verified, works with Claude Code, Cursor, and more.

Security review

✓ Passed

No issues found. Passed automated security review. · v0.1.0 How review works →

  • Prompt-injection patterns
  • Secret / credential exfiltration
  • Dangerous shell & filesystem operations
  • Untrusted network calls
  • Known-malicious package signatures

What it can access

  • Network access Used
  • Filesystem access No
  • Shell / process execution No
  • Environment & secrets No
  • Dynamic code execution No

From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.

View the full security report →

Verified badge

Passed review? Show it. Paste this badge into your README, it links to the public security report.

AgentStack Verified badge Links to your public security report.
[![AgentStack Verified](https://agentstack.voostack.com/badges/verified.svg)](https://agentstack.voostack.com/security/report/skill-opendatahub-io-ai-helpers-coderabbit-review)

Reliability & compatibility

Security review passed
0 installs to date
no reviews yet
2mo ago

Declared compatibility

Claude CodeClaude Desktop

Compatibility is declared by the source manifest. End-to-end runtime verification is coming, see below.

Preview Execution monitoring

We're building live execution health for every listing: tool-call success rate, median latency, uptime, and last-checked timestamps, measured, not self-reported. It isn't live yet, so we don't show numbers we can't stand behind.

How agent discovery & health will work →
Are you the author of Coderabbit Review? Claim this listing to set pricing, connect Stripe payouts, and keep 70% of every sale.
Sign up to claim

About

CodeRabbit PR Review Handler

Fetch CodeRabbit comments from a GitHub PR, evaluate each one, and take action: apply a code fix or post a reply.

PR: $ARGUMENTS

Step 1: Resolve the PR and Repository

Determine the owner, repo, and pullNumber:

  • If $ARGUMENTS is a PR number, use it directly.
  • Otherwise, detect the PR from the current branch using gh pr view or gh pr list.
  • Use gh repo view --json owner,name to get the upstream coordinates. If that fails (e.g., no default remote set), fall back to parsing git remote -v to identify the upstream GitHub repository.
  • The repo may be a fork with multiple remotes. Do not assume names like upstream/origin. Prefer repo coordinates from PR metadata (gh pr view --json), then gh repo view, then git remote -v. If multiple candidates remain, ask the user.
  • If you cannot determine the PR automatically, ask the user.

Step 2: Fetch CodeRabbit Comments

Inline review comments (CodeRabbit's line-level suggestions):

gh api --paginate repos///pulls//comments

PR-level comments (CodeRabbit's summary/walkthrough):

gh api --paginate repos///issues//comments

Filter both lists to only comments where user.login equals coderabbitai[bot].

For PR-level comments, skip any comment whose body contains ` or [If fix: show the exact code diff to apply]

[If reply or dismiss: show the draft reply text]


After presenting ALL comments, wait for the user to respond. The user may want to:

  • Discuss — ask questions, understand a comment better, debate whether it's valid. Act as a knowledgeable colleague: explain the trade-offs, share context, help them form their own opinion. Stay in this discussion mode as long as the user is engaging. Do NOT push toward action or present menus while the user is still exploring.
  • Act — when the user indicates they're ready (e.g., "let's fix these", "go ahead", "apply #1 and #3"), proceed with the requested actions.
  • Decide per-comment — if the user wants to go through comments one by one, walk through each and ask what to do.

When the user is ready to act, the available actions per comment are:

  • Apply fix — apply the code change (only when a fix is proposed)
  • Post reply — post the draft reply to the PR
  • Edit & post — let the user adjust the reply text first, then post
  • Skip — do nothing with this comment

Critical rule: Never apply fixes or post replies without explicit user approval. If unsure whether the user wants to act or keep discussing, keep discussing.

Step 5: Execute Actions

For "Apply fix":

  1. If the fix involves externally-sourced values (SHA256 hashes, version strings, URLs suggested by CodeRabbit), verify them independently before applying:
  • For checksums: download the artifact and run sha256sum to confirm
  • For version/URL claims: fetch only from known allowlisted HTTPS domains (e.g., github.com, pypi.org); reject localhost, link-local, and private IP ranges. Use hardened curl flags: --proto '=https' --tlsv1.2 --fail --location --max-time 15 --connect-timeout 5
  • CodeRabbit uses web search internally and can return incorrect values
  1. Apply the change using the Edit tool
  2. Show a confirmation of what changed

For "Post reply" (inline review comment):

gh api -X POST repos///pulls//comments//replies -f body=""

For "Post reply" (PR-level comment):

gh api -X POST repos///issues//comments -f body=""

For "Edit & post": Use AskUserQuestion to ask the user to provide their edited reply text, then post it using the appropriate gh api call above.

For "Skip": Move on.

Step 6: Summary

After all comments are processed, output:

CodeRabbit Review — Done

Actions taken:

  • Fixed: N comment(s) — code changes applied
  • Replied: N comment(s) — replies posted to PR
  • Dismissed: N comment(s) — brief acknowledgements posted
  • Skipped: N comment(s)

Next steps:

  • If code was changed: do NOT lump all changes into a single generic commit. Group related changes into separate logical commits, each with a message that explains why the change was made (not just "address CodeRabbit feedback"). Unrelated fixes from different CodeRabbit comments should be separate commits. Push to the PR head branch (fork or same-repo, whichever feeds the PR). Use gh pr view --json headRefName,headRepositoryOwner plus git remote -v and git branch -a to confirm destination before pushing.
  • If only replies were posted: CodeRabbit may re-evaluate on re-review trigger

Notes

  • CodeRabbit bot login is exactly coderabbitai[bot] — filter by this
  • For inline review comments: key fields are id, path, line, original_line, diff_hunk, body
  • For PR-level comments: key fields are id, body, user.login
  • Be professional and concise in replies — acknowledge valid points, politely explain disagreements
  • For nitpicks: "Thanks, noted — will address in a cleanup pass" is often the right reply
  • NEVER post a reply or edit a file without explicit user approval

Gotchas

  • CodeRabbit's bot login is exactly coderabbitai[bot] — using any other string (e.g., coderabbit) will silently miss all comments.
  • Walkthrough and summary posts look like actionable comments but must be filtered out by checking for `` markers.
  • Externally-sourced values in CodeRabbit suggestions (SHA256 hashes, version strings) can be wrong — always verify independently before applying.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

Install and usage instructions live in the source repository linked above.

Reviews

No reviews yet, be the first.

Versions

  • v0.1.0 Imported from the upstream source.