Review Feedback Support

Once a PR is open, FlightDesk can aggregate feedback from the review tools you use — GitHub reviewers, SonarQube, GitHub Copilot — and surface it in a single Checks panel. These integrations support the workflow; the core task remains the shared team view, preview URL, and QA handoff.

The Checks Panel

Each active integration appears as a check on the task detail page:

• Passed — no issues found, this check is green
• Pending — issues have been found that need attention
• Failed — critical issues that should block merge (e.g. security hotspots in SonarQube)

Checks update in real time as integrations report back.

Supported Integrations

GitHub PR Reviews

FlightDesk tracks all reviews left on the PR — human reviewers, bots, and GitHub Copilot suggestions. Each reviewer appears as a separate check showing their approval status and any comments.

Setup: Automatic once GitHub is connected. No additional configuration needed.

GitHub Copilot

If Copilot code review is enabled on your repository, its suggestions are fetched and surfaced as a check. Copilot suggestions are fetched via FlightDesk's proxy so Claude Code can read them with proper authentication.

Setup: Enable Copilot code review in your GitHub repository settings. FlightDesk picks it up automatically.

SonarQube

SonarQube scans appear as a check showing issue counts by severity. FlightDesk receives scan results via webhook and links to the SonarQube project for full details.

Setup: See SonarQube Setup.

The Review Prompt

When there are pending checks, the task's Prompts tab offers a review prompt to copy — or flightdesk prompt <task-id> from the terminal. It generates a prompt that:

1. Lists every pending issue from every integration
2. Provides authenticated API endpoints to fetch the full content of each issue
3. Instructs Claude Code to fix each issue and mark it resolved via the API
4. Includes a task token so Claude Code can authenticate against FlightDesk's proxy

Paste this prompt into Claude Code. It will:

• Call FlightDesk's proxy to fetch the actual comment or issue text
• Make the code changes needed to address each piece of feedback
• Push the changes
• Call the API to mark issues as addressed

After the push, integrations re-run automatically (SonarQube re-scans, Copilot re-reviews) and the checks panel updates.

The Proxy API

FlightDesk proxies requests to GitHub and SonarQube on behalf of Claude Code. This means:

• Claude Code doesn't need GitHub or SonarQube credentials
• The task token embedded in the review prompt is scoped to that task only
• Tokens expire automatically — Project Settings → General → Token TTL (hours), 72 by default

Refreshing the Review State by Hand

Review state normally arrives by webhook. To pull it now instead of waiting:

flightdesk task sync <task-id>

That re-reads the pull request's Copilot feedback, Claude reviews and human reviews from GitHub.

Iterating Until Green

The typical review loop:

1. PR opens → checks appear in FlightDesk
2. Issues found → copy the review prompt from the Prompts tab
3. Paste into Claude Code → it fixes the issues and pushes
4. Re-checks run automatically
5. Repeat until all checks are green
6. A person reviews the final diff on the preview — the QA phase — and approves
7. Merge

On an orchestrated project you do not run this loop by hand: a failing check or an unresolved review thread on work with no live session dispatches a fix turn, which does the same thing and ends waiting on CI again. See Agents and Orchestration.