[joshschmelzle]

Summary

Adds an impact gate for review findings. A reviewer reports a finding only with a concrete user scenario, or for data loss, a security hole, or a crash, and labels it Blocker, Follow-up, or Drop.

Testing

Documentation only. No code or CI behavior changes.

---

Note

Low Risk
Documentation-only guidance in AGENTS.md; no code, CI, or runtime behavior changes.

Overview
Updates agent guidance in agents/AGENTS.md so builders and reviewers stay aligned on scope and noise.

Under How to work, adds a rule to avoid designing for rare or hypothetical futures that were not requested.

Adds a Reviewing section (for agent reviewers, QA, and the existing self-review step). Findings must start with a concrete user scenario or qualify as data loss, security, or crash risk; otherwise they are dropped. Each reported finding gets a single label: Blocker, Follow-up, or Drop, with an explicit floor for serious bugs. Style, naming, docs-only nits, and unrequested edge cases are out of scope; empty reviews are OK.

^{Reviewed by Cursor Bugbot for commit 315792a. Bugbot is set up for automated code reviews on this repo. Configure here.}

[cursor]

Approved. Cursor Bugbot completed successfully with no findings that need human review, and no approval policy requires extra review on this documentation change.

  

_{Sent by Cursor Approval Agent: Pull Request Router and Approver}