Prevent output injection in publish-release workflow

[alessandrobologna]

Motivation

• The publish-release workflow wrote a user-controlled release_tag input directly to $GITHUB_OUTPUT, allowing newline-based output injection to override checkout_ref and version and cause releases to be built from unintended refs.
• This could let an actor with workflow_dispatch rights cause the workflow to check out and publish code from an attacker-controlled ref while using the workflow's AWS publishing credentials.

Description

• Add an explicit validation to reject release_tag values containing newline characters (\n or \r) in the Resolve release ref step of .github/workflows/publish-release.yml to block output injection.
• Replace echo-based writes to $GITHUB_OUTPUT with printf-based writes for tag, version, and checkout_ref to ensure literal, single-line output assignments.

Testing

• Inspected the workflow file with sed -n '1,140p' .github/workflows/publish-release.yml to verify the new newline validation and that the printf writes are present.
• Examined the relevant lines with nl -ba .github/workflows/publish-release.yml | sed -n '48,86p' to confirm the Resolve release ref logic now rejects \n/\r and writes outputs with printf.
• Confirmed the change is limited to release-ref output handling and no other workflow behavior was modified; all inspections succeeded.

---

Codex Task

[Copilot]

Pull request overview

This PR hardens the release publishing workflow against $GITHUB_OUTPUT injection from manually supplied release tags.

Changes:

• Rejects release_tag values containing newline or carriage-return characters.
• Replaces echo output writes for release ref outputs with printf single-line assignments.

---

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