[Follow-up] Address verification concerns from PR #697

[github-actions]

Follow-up from PR #697

Original PR: #697 - chore(codex): bootstrap PR for issue #692
Verification Verdict: Unknown

---

Concerns from Verification

• Acceptance criteria are fundamentally about behavior in a consumer repo (“detects and links duplicate” and “allows unique issues through”), but this PR mainly adds a smoke-test script and unit tests; it does not include any workflow changes or an executable, repo-verified artifact demonstrating linking actually occurs. Code can only infer/observe via API responses; it cannot guarantee the workflow ran and performed linking.
  • The smoke script appears to be designed to interact with GitHub Issues (create issues, look for dedup results), which introduces risk of flaky behavior due to eventual consistency, permissions, rate limits, or differing repo configuration. If the script relies on timing/polling, the implementation may intermittently misclassify outcomes without robust retries/backoff.
  • Large new script (≈588 LOC) suggests substantial logic; without seeing strong decomposition and clear separation of concerns (API client vs. domain checks vs. CLI), maintenance could be difficult. The PR adds tests, but the size indicates a higher chance of untested edge paths (e.g., pagination, API errors, partial permissions).
  • Potential security/operational risk if the smoke script can create/modify issues in arbitrary repos with a token; safeguards (dry-run mode, explicit allowlist, clear confirmation prompts, least-privileged scopes) should be ensured. If not present, accidental issue spam is possible.
  • Completeness relative to scope (“Test the agents-dedup.yml workflow with real issues in a consumer repo”): no repo configuration/fixtures or documented invocation evidence is added beyond script/test code; the PR updates pr_body.md and adds an agent marker file, but does not encode the actual end-to-end validation as a reproducible check against a real consumer repo.

Suggested Tasks

• Review the concerns identified above
• Address each issue or document why it is not applicable
• Update tests if needed
• Consider re-verification after changes

---

Context

This issue was created from verification feedback on a merged PR.

How to use this issue

1. Add agents:optimize label to get AI-suggested improvements
2. Add agents:apply-suggestions to format for agent work
3. Add agent:codex to assign to an agent

Or work on it manually - the choice is yours!

---

Auto-generated by verify-to-issue workflow