mcpField
C
cli-review

cli-review

github.com/greptileaiTesting & quality428593 installsUpdated 2d ago
Repository

About cli-review

A skill for running Greptile CLI reviews from a local Git checkout and turning results into an actionable summary. It guides repository checks, authentication, review execution, and failure handling.

Key features

  1. 01Checks Git repository context
  2. 02Verifies Greptile CLI and authentication
  3. 03Runs JSON review with text fallback
  4. 04Prioritizes findings and affected files
  5. 05Reports failures and next actions

Use cases

  1. 01Review changes in a local checkout
  2. 02Summarize review findings by severity
  3. 03Diagnose review command or login failures

Install

Copy the command and run it in a terminal.

$npx skills add https://github.com/greptileai/skills --skill cli-review

After installing, type/cli-reviewin Claude Code to confirm it loaded.

Details

From the project README

CLI Review

Run a Greptile review from the local checkout and summarize the findings.

Instructions

1. Confirm repository context

Start from the current repository root:

git rev-parse --show-toplevel

If the command fails, tell the user that the Greptile CLI review must be run from a git repository.

2. Check for the Greptile CLI

Check whether greptile is installed:

command -v greptile

If it is missing, do not install it automatically. Ask the user for permission, then show the recommended install command:

npm i -g greptile

If npm is unavailable, offer the shell installer fallback:

curl -fsSL "https://greptile.com/cli/install" | sh

After installation, re-run command -v greptile.

3. Ensure authentication

Check the signed-in account:

greptile whoami

If the CLI reports that authentication is missing, run:

greptile login

Wait for the user to complete the login flow before continuing.

4. Run the review

Prefer JSON output:

greptile review --json

If JSON output is unsupported or fails with a usage error, fall back to:

greptile review --agent

Do not hide the raw command failure if both commands fail. Summarize the failing command and the next action the user needs to take.

5. Summarize results

Parse JSON output when available and report:

  • Review status
  • Number of findings
  • Highest severity findings first
  • Files that need edits
  • Suggested next command or fix path

When output is plain text, preserve the same structure as much as possible. Keep the summary concise and focused on actionable findings.

Comments

Markdown supported