[bens-agent]

Before: the AI review ran whatever was on mirurobotics/.github main, so a push there changed this repo's review without any change here.

After: the caller pins the reusable workflow to commit 78e2fba, which has the same content as main today. A future workflow change reaches this repo only through a PR here that bumps the SHA.

How: the uses: ref changes from @main to the full SHA, and a header comment explains why. This PR's own review also gives the repo its first completed review, so ai-review can be made required next.

🤖 Generated with Claude Code

https://claude.ai/code/session_011yoL2D46V9c74Q9aJXgXZu

---

Generated by Claude Code

---

^{Need help on this PR? Tag @codesmith-bot with what you need. Autofix is disabled.}

[chatgpt-codex-connector]

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.

[github-actions]

AI review: approve for fedb666

This PR pins the reusable AI review workflow in .github/workflows/ai-review.yml from @main to a full commit SHA (78e2fba…) and adds a header comment explaining why. This is a CI/change-management hardening step. Nothing else changes: permissions, triggers and secrets are the same. I couldn't check that the SHA exists in mirurobotics/.github or matches its main, because the gh api call needed approval that wasn't given. Because the workflow uses pull_request_target, this PR's own run still uses @main, so a bad SHA would only show up on the next PR, as a failed required check. I left a non-blocking inline note asking the author to confirm the SHA. No blocking issues.

Review run