Skip to content

Setup Qoder AI Workflow - #2

Merged
Mount4in merged 2 commits into
mainfrom
qoder-setup-1777017797
Apr 24, 2026
Merged

Setup Qoder AI Workflow#2
Mount4in merged 2 commits into
mainfrom
qoder-setup-1777017797

Conversation

@Mount4in

Copy link
Copy Markdown
Owner

Setup Qoder AI workflow

This PR adds GitHub Actions workflows to integrate Qoder.

Included files:

  • .github/workflows/qoder-auto-review.yml
  • .github/workflows/qoder-assistant.yml

Usage:

  • Open or update a Pull Request to trigger automated review
  • Mention @qoder in issues or PR comments to get assistance

@qoderai qoderai Bot left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

👋 Review Summary

This PR cleanly introduces two focused GitHub Actions workflows to integrate Qoder for comment-driven assistance and automatic PR review, with clear triggers and minimal, easy-to-reason-about steps. The overall intent and structure look good for enabling AI-assisted workflows in this repo.

🛡️ Key Risks & Issues

  • Command injection risk in qoder-assistant Build Arguments step: In .github/workflows/qoder-assistant.yml, the Build Arguments step builds a double-quoted Bash string ARGS that embeds github.event.comment.body and other comment-derived fields directly into the run script. Because comments are untrusted input, a crafted comment containing quotes, backticks, or $(...) could break out of the string and cause arbitrary commands to run in the job, with access to QODER_PERSONAL_ACCESS_TOKEN and write permissions on issues/PRs. This is a high-priority security issue; we should restructure how we pass comment data (e.g., via ${{ toJson(...) }} into env or outputs, or by letting the Qoder action receive raw fields as inputs) so the shell never interprets user content as code.
  • Overly broad id-token permissions: Both workflows (qoder-assistant and qoder-review) request id-token: write in their permissions blocks, but there is no obvious use of OIDC tokens in the job definitions. Keeping id-token enabled without a concrete need increases the blast radius if the job is compromised, since an attacker could mint OIDC tokens to talk to external systems. Unless the Qoder action specifically requires OIDC here, it would be safer to drop id-token from these workflows to align with least-privilege best practices.

🧪 Verification Advice

  • For the assistant workflow, exercise comments that include @qoder plus various special characters (quotes, backticks, $(...), and multi-line markdown) in a test repo or branch, and confirm that the job still runs successfully and that the generated prompt is well-formed once the interpolation pattern is fixed.
  • For both workflows, verify behavior when QODER_PERSONAL_ACCESS_TOKEN is missing or invalid and when permissions (issues/pull-requests) are reduced, so that failure modes are clear and documented for future operators.

💡 Thoughts & Suggestions

  • The trigger scoping and avoidance of running on every event are well thought out; the condition to ignore [bot] comments is a nice touch to prevent loops. After addressing the command injection concern and tightening permissions, these workflows should provide a solid and maintainable foundation for Qoder-based assistance and reviews in this repo.

🤖 Generated by QoderView workflow run

Comment thread .github/workflows/qoder-assistant.yml
@Mount4in
Mount4in merged commit 537e0c8 into main Apr 24, 2026
4 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant