agent-reviews

Manage GitHub PR review comments from the terminal and from AI coding agents.

PR review bots (Copilot, Cursor Bugbot, CodeRabbit, etc.) leave inline comments on your pull requests. agent-reviews gives you a CLI to list, filter, reply to, and watch those comments, plus a Claude Code skill that automates the entire triage-fix-reply loop.

Why

gh CLI is fragile for review comments. Agents frequently get the syntax wrong, fail to paginate, and can't reliably detect whether a comment has been replied to. agent-reviews provides a single, purpose-built interface that handles all of this correctly.

Bot reviews create a doom loop. You fix one round of findings, push, and new comments appear. Fix those, push again, more comments. This cycle can eat hours. The included skill solves it with an integrated watcher that keeps fixing and replying until the bots go quiet.

Works in Claude Code for the web. Most solutions rely on local tooling that isn't available in cloud environments. agent-reviews works everywhere Claude Code runs, so you can kick off a session in your browser, let it resolve all bot findings autonomously, and come back to a clean PR.

Install

CLI (npm)

npm install -g agent-reviews

Agent Skill

Installs the full automated workflow as a slash command (no npm install required).

Agent-agnostic (works with any agent that supports Agent Skills):

npx skills add pbakaus/agent-reviews@agent-reviews

Claude Code plugin:

/plugin marketplace add pbakaus/agent-reviews
/plugin install agent-reviews@agent-reviews

Both register the /agent-reviews command. The skill bundles its own scripts, so nothing is downloaded at runtime.

You can also use both: install the CLI globally for direct terminal use, and the skill for the agent workflow.

Authentication

The primary authentication method is the GitHub CLI. If you're logged in with gh auth login, agent-reviews picks up the token automatically. No configuration needed.

For environments where gh isn't available (like Claude Code Cloud, which routes git through an HTTPS proxy), agent-reviews falls back to:

1. GITHUB_TOKEN environment variable
2. .env.local in the repo root

The proxy environment is also why agent-reviews includes undici ProxyAgent support. When HTTPS_PROXY is set, GitHub API requests are routed through it automatically.

Resolution order (first match wins):

1. GITHUB_TOKEN environment variable
2. .env.local in the repo root
3. gh auth token (GitHub CLI)

CLI Usage

# List all review comments on the current branch's PR
agent-reviews

# Only unresolved comments
agent-reviews --unresolved

# Only unanswered bot comments
agent-reviews --unanswered --bots-only

# Full detail for a specific comment (diff hunk + replies)
agent-reviews --detail 12345678

# Reply to a comment
agent-reviews --reply 12345678 "Fixed in abc1234"

# JSON output for scripting / AI agents
agent-reviews --json

# Watch for new comments (polls every 30s, exits after 10 min idle)
agent-reviews --watch --bots-only

# Target a specific PR (otherwise auto-detects from branch)
agent-reviews --pr 42

Options

Claude Code Skill

The /agent-reviews skill automates the full PR review bot workflow:

1. Fetch all unanswered bot comments via npx agent-reviews --bots-only --json
2. Evaluate each finding (true positive, false positive, or uncertain)
3. Fix real bugs and run lint/type-check
4. Dismiss false positives with an explanation
5. Reply to every comment with the outcome
6. Watch for new comments as bots re-analyze the fixes
7. Report a summary of all actions taken

Skill behavior

• True positives get fixed, verified, and replied with Fixed in {commit}
• False positives get replied with Won't fix: {reason}
• Uncertain findings prompt the user via AskUserQuestion
• All fixes are batched into a single commit before polling begins
• Watch mode runs for up to 10 minutes, processing any new findings

How It Works

Comment types

agent-reviews fetches three types of GitHub PR comments:

Meta-comment filtering

Status updates from bots are automatically filtered out:

• Vercel deployment status ([vc]:...)
• Supabase branch status ([supa]:...)
• Cursor Bugbot summary ("Cursor Bugbot has reviewed your changes...")

Only actionable findings are shown.

Reply status

Each comment displays its reply status:

Watch mode

Polls the GitHub API at a configurable interval and reports new comments as they appear. Outputs both formatted text and JSON for AI agent consumption. Exits automatically after a configurable inactivity timeout (default: 10 minutes).

FAQ

How can I make the skill review human comments as well?

The skill defaults to --bots-only because that's the most common use case, but it's easy to change. Open your local copy of SKILL.md (inside skills/agent-reviews/) and remove the --bots-only flag from the commands in Steps 2, 6, and anywhere else it appears. The skill will then fetch and process all comments, regardless of author.

License

MIT