Menu
Use when performing a holistic code review of an implementation - routes to codex:rescue or falls back to a direct subagent, both running the house-style:holistic-review skill
Instalar com Codex ou Claude Copie este prompt, cole no Codex, Claude ou outro assistente e deixe que ele revise a página da skill e instale para você.
Baseado na classificação ocupacional SOC
| name | holistic-review |
| description | Use when performing a holistic code review of an implementation - routes to codex:rescue or falls back to a direct subagent, both running the house-style:holistic-review skill |
| context | fork |
| allowed-tools | Bash(node *), Bash(find *), Read(*) |
Perform a holistic code review by routing to the best available review backend. The review methodology itself lives in the house-style:holistic-review skill -- this skill only decides where it runs. Prefers codex:rescue when available; falls back to a direct subagent.
You should have two pieces of information before invoking this skill:
If the plan path is missing, stop and ask. If the base branch is missing, the reviewer will resolve it (origin/HEAD, else main/master) -- pass it through as "not provided".
This skill requires the house-style plugin. Scan your available skills list for house-style:holistic-review. If it is not present, stop and report that the house-style plugin must be installed (/plugin install house-style@llm-plugins).
Before doing anything else, determine whether codex:rescue is available.
Scan your available skills/commands list (shown in your system context) for codex:rescue. Record the result explicitly in your response:
codex:rescue: AVAILABLE / NOT AVAILABLE
You MUST write this line before proceeding. Do not skip it.
This is the primary path. You may not skip it when codex:rescue is available. Falling through to the fallback path when codex:rescue is detected is a failure.
Resolve the fully qualified path of the house-style skill file:
find ~/.claude/plugins -path '*house-style*skills/holistic-review/SKILL.md' -print -quit
If nothing is found, fall back to the subagent path below.
Invoke SkillTool("/codex:rescue --fresh --foreground "{prompt}"). Pass a prompt (the {prompt} variable) containing all three of:
You are performing a holistic code review. Follow this skill definition exactly. READ THIS SKILL DEFINITION: {paste the FULLY QUALIFIED house-style holistic-review SKILL.md path here}
Review Parameters:
- Implementation plan: {absolute_plan_path}
- Base branch: {base_branch}
Spin up a general-purpose subagent with a clean context. It must load the house-style skill and follow it exactly.
<invoke name="Task">
<parameter name="subagent_type">general-purpose</parameter>
<parameter name="description">Holistic code review</parameter>
<parameter name="prompt">
You are performing a holistic code review. Invoke the Skill tool with `house-style:holistic-review` and follow that skill exactly, including its output contract. Do not fix any code; report findings only.
Review Parameters:
- Implementation plan: {absolute_plan_path}
- Base branch: {base_branch}
Return the full review report as your final message.
</parameter>
</invoke>
MANDATORY: Address ALL issues from the holistic code review. Do not complete until all issues have been addressed. Loop back through Routing after fixing until you get a clean (PASS) review.
| Mistake | Fix |
|---|---|
| Skipping detection and going straight to the subagent fallback | Always record the detection result first. The gate is mandatory. |
| Running the review inline in this fork instead of routing it | This context carries implementation bias. Always route to codex or a clean subagent. |
Forgetting --fresh on codex:rescue | Without it, Codex may resume a stale thread. Always pass --fresh. |
Use when reviewing a completed implementation holistically - a finished branch, a completed plan step, or a diff before PR. Reviews every changed line against the implementation plan (if one exists) and high engineering standards across seven dimensions, verifies tests by deliberately breaking crucial code, and emits severity-classified findings with a PASS/FAIL verdict.
ALWAYS use before any implementation work, review, or planning: designing, writing, editing, refactoring, debugging, reviewing, or generating source code, tests, scripts, config-as-code, or behavior-affecting files. This is a baseline engineering skill and should be invoked even when a more specific language, framework, testing, or database skill also applies.
Use when asked to write, draft, file, or open a GitHub issue - mandates exploring the relevant code to understand the whole problem space, reasoning any proposed solution, and structuring the issue with summary, what/why, references, nuances, open questions, and acceptance criteria. Adds professionalism and no-sensitive-info discipline for public repos.
Use when creating, editing, triaging, or cleaning up GitHub issues - a mechanical pass that verifies the title is accurate and current, labels/assignees/milestone are correct, and the body is valid, well-formed markdown, all via the gh CLI. Does not judge content quality.
Use when running Python, installing Python packages, creating virtualenvs, or executing Python CLI tools - mandates uv/uvx over direct python/pip/pipx/venv calls when uv is installed, and gives the drop-in command mappings
Use when the user provides a reference implementation (URL, repo path, branch, file) plus a verb like "use", "copy", "adapt", "port", "model after" - mandates fetching/reading the reference before designing, and lifting the structure rather than reimagining it.