[utsab345]

Fixes #4424. Adds pr_code_suggestions.publish_failure_comment to control failure comments from /improve without changing command failure status. The default preserves existing behavior, while disabled comments still remove progress output and record failures. Documentation and lifecycle tests are included.

[the-pr-agent-bot]

PR Reviewer Guide 🔍

Here are some key observations to aid the review process:

⏱️ Estimated effort to review: 2 🔵🔵⚪⚪⚪

🧪 PR contains tests

🔒 No security concerns identified

⚡ Recommended focus areas for review Configuration count Adding this active setting increases the configuration-key count, but the hard-coded count in tests/unittest/test_config_reference.py is not updated. That test will fail even though the generated reference includes the new setting. Update the expected count alongside this addition. publish_failure_comment=true

---

⚙️ Agent run details

• Model: openai/gpt-6.1-sol
• Tokens: 6,286 in / 236 out / 6,522 total
• Time cost: 10.8s
• AI calls: 1

[IsmaelMartinez]

Thanks @utsab345 for taking #4424 and keeping the source change to one line, matching the issue and the /review key, approving and merging.