菜单
Board of Judges v2 — tool-grounded, eval-backed panel code review. Shells out to the `boj` Python CLI. Usage: /judge <file> [--panel X] [--solo X] [--all]
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
| name | judge |
| description | Board of Judges v2 — tool-grounded, eval-backed panel code review. Shells out to the `boj` Python CLI. Usage: /judge <file> [--panel X] [--solo X] [--all] |
When the user invokes /judge <file>:
Verify boj is on PATH with boj version. If missing, instruct them to
uv sync --dev or pip install -e . from the repo root.
Run:
boj judge <file> [--panel <tags>] [--solo <judge-id>] [--all] \
--markdown judgements/gemini-<timestamp>-<slug>.md \
--output judgements/gemini-<timestamp>-<slug>.json
Read the markdown file and display it verbatim. Do not editorialize.
If the exit code is non-zero and the verdict was FAIL, summarize the
critical issues in one line at the top.
If the user passes --post-to-pr, check for an active PR context via
gh pr view --json number and post the markdown as a comment.
The actual review logic (judge selection, tool grounding, verification,
synthesis) is owned by boj. This skill is a thin launcher.
Board of Judges v2 — tool-grounded, eval-backed panel code review. Invokes the `boj` CLI; returns a structured verdict report. Usage: /judge-v2 <target> [--panel X] [--solo X] [--all] [--post-to-pr]
Generates all Claude Code judge agent files from role stubs. Run once to populate .claude/agents/ with 87+ specialist judge personas. Usage: /bootstrap-judges [--dry-run] [--review <agent-filename>]
Board of Judges coordinator. Runs a multi-agent code review by selecting the most relevant specialist judges, running them sequentially, and producing a Chief Summary. Invoke as: /judge <target> [--panel X] [--solo X] [--all] [--post-to-pr]
Lists all available Board of Judges agents, grouped by section, with their tags and tier. Use to find the right --solo or --panel target for /judge.