agent-reviews skill
A CLI and three skills that work through review-bot comments on a PR: assess findings, fix code, reply and wait for checks
- 306 ★ on the repository
- Last commit
- MIT
Broad access
Why this level
- By default the skill edits code, commits, pushes to your branch and writes replies under your GitHub account
- The tool reads a token from `GITHUB_TOKEN`, `GH_TOKEN`, `.env.local` or `gh auth token`
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/agent-reviewsDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- By default the skill edits code, commits, pushes to your branch and writes replies under your GitHub account
- The tool reads a token from `GITHUB_TOKEN`, `GH_TOKEN`, `.env.local` or `gh auth token`
This is third-party code. Review the repository files before installing.
What it does
agent-reviews addresses the endless loop with review bots: you fix one round of comments from Copilot, Cursor Bugbot or CodeRabbit, push, and the bot leaves new ones. The tool is a `agent-reviews` CLI plus three skills. The CLI lists, filters and replies to review comments on the current PR and watches for new ones. The skills run the whole cycle for the agent. The `resolve-agent-reviews` skill takes unanswered bot comments and judges each: a real bug, a false positive or an unclear case it asks you about. It fixes real bugs, runs lint and type checks and batches the fixes into one commit. The agent then replies to every comment with `Fixed in {commit}` or `Won't fix: {reason}`. After that the skill waits for the head commit's checks and repeats until the PR is clean. The other two skills work the same way: `resolve-reviews` handles all comments, `resolve-human-reviews` only human ones.
Who it is for. For developers whose PRs are reviewed by Copilot, Cursor Bugbot, CodeRabbit and other bots, and teams who want to hand the routine of answering comments to an agent.
Good fit when
- A bot left a dozen comments on a PR, and you need to separate real bugs from false positives and reply to each
- New bot comments appear after every push, and you want to close the fix loop automatically
- The agent runs in a cloud environment or Codespaces: the tool talks to the GitHub API with a token and does not depend on local utilities
- You need review comments as JSON for a script: the `--json` flag with `--unresolved` and `--bots-only` filters
- You want to wait for the head commit's checks and see failed checks together with unanswered comments: `--wait-checks`
Not a fit when
- You need a review of the code itself: the tool works through other people's PR comments and does not check code on its own
- You cannot trust an agent with pushing fixes: the skills allow `git add`, `git commit` and `git push` on your branch
- The comments concern architecture and business logic: in such cases the skill asks you rather than deciding
- The repository is not on GitHub: the tool uses the GitHub API, and for Enterprise the address changes through `GITHUB_API_URL`
Example request
Work through the bot comments on this PR: fix real bugs, reply to the rest and wait for green checksLimitations
You need git, `gh` (GitHub CLI) and Node.js, as the skill's metadata states. GitHub access is taken in order from `GITHUB_TOKEN`, `GH_TOKEN`, a `.env.local` file in the repository root and the `gh auth token` command. In cloud and proxy environments `HTTPS_PROXY` is supported.
The PR is detected from the branch, and the lookup order for the head repository on forks is documented separately. If several PRs match or the head repository is inaccessible, the tool asks for an explicit target through `--pr` or `GH_REPO`. Service comments such as Vercel deploy status or CodeRabbit and Copilot summaries are filtered out. The skill batches fixes into one commit and pushes it, so on a push-protected branch it needs separate setup.
How to disable. Remove a skill installed through `npx skills` from the agent's skills folder, and uninstall the global CLI with `npm uninstall -g agent-reviews`. Nothing else is configured: the token is read from environment variables.
FAQ
How do I install agent-reviews in Claude Code?
For the skill, run `npx skills add pbakaus/agent-reviews@resolve-agent-reviews`. You can swap `resolve-agent-reviews` for `resolve-reviews` or `resolve-human-reviews`. The skills call `npx agent-reviews` on the fly, so no separate `npm install` is needed.
How do the three skills differ?
`resolve-reviews` handles all comments, human and bot. `resolve-agent-reviews` takes only bot comments such as Copilot and Cursor. `resolve-human-reviews` takes only human comments and watches until reviewers go quiet.
What does the agent reply to comments?
For real bugs and sound feedback it replies `Fixed in {commit}`, and for false positives it writes `Won't fix: {reason}`. If a case is unclear, the skill asks you. All fixes land in one commit before the waiting starts.
How do I use the CLI without an agent?
Install it with `npm install -g agent-reviews`. The `agent-reviews` command lists comments on the current branch's PR, `--unresolved` keeps unresolved ones, `--reply 12345678 "Fixed in abc1234"` replies, and `--watch` polls for new comments every 30 seconds and exits after 10 minutes of quiet.
How do I hide service comments?
The `--ignore-author 'github-actions[bot]'` flag hides an author and works with `--watch`. Repeat it for several authors; case does not matter. Nobody is excluded by default, while deploy statuses and summaries from known bots are filtered separately.
Discussion
No questions or reviews yet.
Related
SkillsA code-review CLI and plugin for agents: reads a git diff and writes line-level comments with context
An autonomous iteration skill for agents: modify, verify, revert failed changes, and repeat the loop toward a target metric
A Claude Code and Codex plugin with skills for code research, tests, refactoring and independent review of finished changes
A skill that draws every pull request as an architecture and data-flow diagram right in the PR description, to speed up review
A code review skill across many languages and frameworks: architecture, performance, security, anti-patterns and common bugs
A skill for second-agent review of plans or code: Claude Code starts rounds, Codex reviews, and they exchange files
Comparison with similar tools
Entries that share the main search query, and how they differ.
| Entry | Official | Repository stars | Main difference |
|---|---|---|---|
| agent-reviews (this entry) | no | 306 | |
| code-review-skill | no | 2,136 | reviews the code itself across languages and frameworks, while this works through review-bot comments on a PR |
| GitHub CLI skills | no | 46,722 | teaches the agent to call gh in general, while this is a narrow loop of answering review comments and tracking checks |
README in short
The description gives three reasons for the tool: `gh` handles review comments poorly, bots create an endless loop of comments, and cloud environments lack local utilities. It then covers installing the CLI and skills, the order of token lookup, fork support and a custom API address. The main part is a CLI reference with a flag table, the skills' workflow, comment types and the list of bots whose service messages are hidden.
README badge
Original SKILL.md
--- name: resolve-agent-reviews description: Resolve PR review bot findings on current PR. Fetches unanswered bot comments, evaluates each finding, fixes real bugs, dismisses false positives, replies to every comment, and waits for the head commit's checks until a pass is clean. license: MIT compatibility: Requires git, gh (GitHub CLI), and Node.js installed. allowed-tools: Bash(npx agent-reviews *) Bash(pnpm dlx agent-reviews *) Bash(yarn dlx agent-reviews *) Bash(bunx agent-reviews *) Bash(git config *) Bash(git add *) Bash(git commit *) Bash(git push *) metadata: author: pbakaus version: "1.2.0" homepage: https://github.com/pbakaus/agent-reviews --- Automatically resolve findings from PR review bots (Copilot, Cursor Bugbot, CodeRabbit, etc.) on the current PR. Uses a two-phase workflow: fix all existing issues, then poll for new ones until bots go quiet. ## Prerequisites All commands below use `npx agent-reviews`. If the project uses a different package manager, substitute the appropriate runner (e.g., `pnpm dlx agent-reviews` for pnpm, `yarn dlx agent-reviews` for Yarn, `bunx agent-reviews` for Bun). Honor the user's package manager preference throughout. **Cloud environments only** (e.g., Codespaces, remote agents): verify git author identity so CI checks can map commits to the user. Run `git config --global --get user.email` and if empty or a placeholder, set it manually. Skip this check in local environments. ## Phase 1: FETCH & FIX (synchronous) ### Step 1: Fetch All Bot Comments (Expanded) Run `npx agent-reviews --bots-only --unanswered --expanded` The CLI auto-detects the current branch, finds the associated PR, and authenticates via `gh` CLI or environment variables. If anything fails (no token, no PR, CLI not installed), it exits with a clear error message. This shows only unanswered bot comments with full detail: complete comment body (no truncation), diff hunk (code context), and all replies. Each comment shows its ID in brackets (e.g., `[12345678]`). If zero comments are returned, print "No unanswered bot comments found" and skip to Phase 2. ### Step 3: Process Each Unanswered Comment For each comment from the expanded output: