Show HN: adamsreview – better multi-agent PR reviews for Claude Code

· design · Source ↗

TLDR

  • Six-command Claude Code plugin running up to seven parallel sub-agent lenses, an auto-fix loop with regression reversion, and persistent JSON review artifacts.

Key Takeaways

  • /adamsreview:review dispatches up to seven parallel sub-agents (correctness, security, UX, etc.), deduplicates findings, runs a cheap-then-deep validation gate, and optionally adds a Codex CLI pass via --ensemble.
  • /adamsreview:fix applies auto-eligible findings in parallel, re-reviews with Opus, reverts regressions, and commits survivors – one combined commit by default or --granular-commits for per-group.
  • /adamsreview:add injects external findings (cloud /ultrareview paste, Opus scan, teammate notes) into the existing artifact, deduped and validated through the same gates.
  • Review state persists under ~/.adams-reviews/<repo-slug>/<branch>/; steps can run days apart; stored outside ~/.claude/ to avoid repeated permission prompts from Claude Code’s sensitive-file gate.
  • Runs on a standard Claude Code Max subscription; unlike /ultrareview, it does not draw from the Extra Usage pool.

Hacker News Comment Review

  • The one visible commenter questions whether the ceremony-to-value ratio is justified, linking to an example PR to let output speak for itself – no consensus yet on whether the multi-phase pipeline reduces noise in practice.

Original | Discuss on HN