Panel Review

Run a parallel code review across multiple local CLI coding agents (codex, claude, opencode) and synthesize their findings. Use this skill whenever the user asks for a "panel review" / "panel-review", "second opinions on this change", "multi-agent review", "ensemble review", a "deep panel review" / "deep review" / "deep review the findings" / "dig deep into the findings" / "verify each finding" (opt-in deep mode where the coordinator dispatches verification subagents for every finding, then synthesizes their evidence, concrete fixes, and fix rationale — see the "Deep mode" section in SKILL.md for the procedure and output shape), asks to "have multiple agents/LLMs review this", "cross-check this with codex/claude/etc", "fan out a code review", or any similar phrasing asking for independent reviews from outside this conversation. Each panelist runs in a fresh non-interactive subprocess with no shared state — that is the point. Do NOT use when the user just wants the current session to review code itself; use the regular code-reviewer agent for that.

mkdir -p ~/.claude/skills/panel-review
curl -L https://claudskills.com/skills/panel-review/SKILL.md \
  -o ~/.claude/skills/panel-review/SKILL.md

@misc{panel-review-2026,
  author    = {catena-labs},
  title     = {Panel Review [Claude Code skill]},
  year      = {2026},
  publisher = {ClaudSkills},
  url       = {https://claudskills.com/skills/panel-review/}
}