PR Review Agent: make failure runs diagnosable and harden transient setup failures

[Copilot]

Recent PR Review Agent failures were non-diagnosable because affected runs had no retrievable logs/jobs, and failures appeared transient in a narrow time window. This PR improves observability for failed runs and reduces setup fragility in the CLI install phase.

• Auth visibility + scope enforcement

  • Adds an explicit early Verify auth scopes step in pr-review.yml.
  • Emits gh auth status output and fails fast if required scopes are missing (repo, read:org), so auth regressions are visible at the top of the run.

• Failure diagnostics artifact

  • Adds a dedicated if: failure() artifact upload step after cascade review.
  • Captures candidate/debug outputs (prs.txt, cascade logs, legacy review logs) from both /tmp and ${{ runner.temp }} paths with 14-day retention.

• Transient npm install resilience

  • Introduces a small retry wrapper for global npm CLI installs used by review engines.
  • Applies retry behavior to Claude/Gemini installs to reduce one-off network/registry failures from failing the entire run.

- name: Upload debug log on failure
  if: failure()
  uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7.0.1
  with:
    name: pr-review-debug-${{ github.run_number }}
    path: |
      prs.txt
      /tmp/cascade/**
      /tmp/review-*.log
      ${{ runner.temp }}/cascade/**
      ${{ runner.temp }}/review-*.log
    if-no-files-found: ignore
    retention-days: 14

[don-petry]

Dev-Lead — rate-limited (intent: fix-bot-comment)

PR: #353
Please re-trigger manually (re-mention @dev-lead) when the rate limit clears — the original request cannot be reconstructed automatically.

[gemini-code-assist]

Warning

You have reached your daily quota limit. Please wait up to 24 hours and I will start processing your requests again!

[don-petry]

Dev-Lead — rate-limited (intent: human)

PR: #353
Please re-trigger manually (re-mention @dev-lead) when the rate limit clears — the original request cannot be reconstructed automatically.

[don-petry]

Note

@don-petry I received your request but all AI engines are currently rate-limited. Please re-mention @dev-lead when the rate limit clears (estimated: unknown) — I cannot reconstruct the original instruction automatically.

[Copilot]

Pull request overview

This PR improves the diagnosability and resilience of the PR Review Agent workflow by making authentication problems visible early, retrying transient npm install failures, and uploading failure-only debug artifacts.

Changes:

• Add an early “Verify auth scopes” step that prints gh auth status output and fails fast when required scopes are missing.
• Add a small npm_install_retry wrapper and use it for Claude/Gemini CLI installs.
• Upload a debug artifact on workflow failure containing candidate lists and cascade/legacy logs from /tmp and ${{ runner.temp }}.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[don-petry]

@copilot - review, analyze, address and resolve the comments. Ensure CI is passing.

[Copilot]

@copilot - review, analyze, address and resolve the comments. Ensure CI is passing.

Addressed and pushed in 7c1af5e. I fixed both actionable workflow review issues: (1) gh auth status output is now captured and echoed even on failure before exiting, and (2) the artifact path literal now contains only real paths (no inline comment entries). I also rechecked CI logs via GitHub Actions MCP: prior failures were from the old commit state, and the latest pushed changes passed local targeted validations plus parallel Code Review/CodeQL.

[don-petry]

Dev-Lead — rate-limited (intent: human-pr)

PR: #353
The retry cron will re-attempt automatically.

[don-petry]

Note

I received your request but all AI engines are currently rate-limited. I'll retry automatically once the rate limit clears.
Rate limit resets at: unknown

[don-petry]

Dev-Lead — rate-limited (intent: fix-bot-comment)

PR: #353
Please re-trigger manually (re-mention @dev-lead) when the rate limit clears — the original request cannot be reconstructed automatically.

[don-petry]

Dev-Lead — rate-limited (intent: human)

PR: #353
Please re-trigger manually (re-mention @dev-lead) when the rate limit clears — the original request cannot be reconstructed automatically.

[don-petry]

Note

@don-petry I received your request but all AI engines are currently rate-limited. Please re-mention @dev-lead when the rate limit clears (estimated: unknown) — I cannot reconstruct the original instruction automatically.

[don-petry]

Dev-Lead — human-pr (no-changes)

No changes were needed for this PR.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 64a7ac5e27

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Make failure artifact name unique per rerun attempt

Using pr-review-debug-${{ github.run_number }} can make this step fail on workflow re-runs, because github.run_number is stable across re-run attempts while upload-artifact@v4+ fails when an artifact with the same name already exists unless overwrite: true is set. In a fail-then-rerun scenario, the second failed attempt may not upload diagnostics at all due to a 409 conflict, which undermines the new observability path this change is adding.

Useful? React with 👍 / 👎.

[don-petry]

Dev-Lead — human-pr (applied)

Changes committed and pushed.

[don-petry]

No description provided.