菜单
Manually run a PHP code-cleanliness review on a file, at higher quality than the automatic background hook. Optional first argument is FULL or CHANGES; the last argument is the PHP file path.
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
Draft a standardized, reviewer-focused pull request description and - only after explicit approval - update the PR. Optional first argument is a PR number (defaults to the current branch's open PR). Optional argument is a Jira issue ID (e.g. ABC-123).
Research and understand a module, folder, or problem before implementing, capturing what you learn as local notes. Use before starting a complex task or to deepen understanding of an area. Optional argument is a module path/name and/or a Jira issue ID (e.g. ABC-123).
Reviews code changes for quality, security, and adherence to project standards. Use when reviewing PRs, commits, or staged changes.
Run a PR readiness check on the current branch before opening or finalizing a pull request. Optional first argument is a Jira issue ID (e.g. ABC-123). Optional second argument is a base branch name.
| name | php-review |
| description | Manually run a PHP code-cleanliness review on a file, at higher quality than the automatic background hook. Optional first argument is FULL or CHANGES; the last argument is the PHP file path. |
| argument-hint | [full|changes] <file> |
| invocation | user |
You dispatch the php-review sub-agent to review one PHP file. You do NOT perform the
review yourself - your job is to parse the arguments and hand off, forcing the stronger
model because this is a deliberate, manual review (the automatic hook runs the same agent
on its cheaper default).
Arguments after /php-review:
FULL or CHANGES (case-insensitive) selects the scope. Optional; defaults to CHANGES.
CHANGES - review only the git-diff hunks (same scope the background hook uses).FULL - review the entire file.If no file path is given, stop and ask the user which file to review.
Invoke the php-review sub-agent via the Task tool with a sonnet model override
(the manual path always runs at the higher-quality model). Pass exactly this as the prompt,
and nothing else - the agent owns all review criteria:
CHANGES scope: review CHANGES in <file>FULL scope: review FULL <file>Pass the sub-agent's report back to the user verbatim. Do not summarize, rewrite, or comment on it.