How do I install the Codereview Family skill?

Run npx skillmds@latest add horizonrobotics/codereview-family in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Codereview Family safe to use?

This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Codereview Family?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Codereview Family free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Codereview Family?

horizonrobotics (@horizonrobotics) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Codereview Family by horizonrobotics

Name: Codereview Family
Author: horizonrobotics

Codereview Family

Structured code review family for explicit heavy review requests and architecture-sensitive review/evaluation requests. Route to PR/MR review, local changeset review, or architecture review based on the review target and intent. Do not use for routine self-checks or casual bug scans.

horizonrobotics 9be007f 4 files · 13.1 KB Updated Sep 17, 2026

File contents

Use this skill family when the user explicitly asks for a heavy review flow, or when the task is already a review/evaluation request and the auto-trigger rules in references/triggering-and-signal.md select a heavy review path.

Read references/triggering-and-signal.md and references/review-depth-and-delegation.md first. When preparing a final report, also use references/report-composition.md.

Then route to the smallest sub-skill that matches the target:

changeset-codereview/ Use for explicit review of a commit, branch diff, staged diff, working tree diff, patch, or local file set.
prmr-codereview/ Use for explicit GitHub PR or GitLab MR review when PR/MR-specific transport, metadata, or comment posting is required.
architecture-review/ Use for explicit architecture review or for review/evaluation requests where layering, contracts, compatibility surfaces, abstractions, or public API design are the main question.

If the target is a PR/MR or local changeset and the user explicitly asks for architecture review in addition to ordinary review dimensions, or if the reviewed scope materially changes layering, ownership boundaries, dependency direction, compatibility/public surfaces, or other architecture-review dimensions, keep prmr-codereview or changeset-codereview as the active skill. For PR/MR reviews, the delegated changeset-codereview workflow owns applying any paired architecture-review/ dimensions. A paired skill is a logical report input and normally shares the same review subagent. In both cases, the active skill owns the main report and must summarize the paired architecture results in Related review inputs.

Do not use this family for:

routine implementation self-checks
quick bug scans during development
ordinary debugging
style-only or taste-only feedback

Use the smallest review surface that satisfies the request. The heavier the workflow, the more explicit the user request should be.

Treat reviewer roles as logical review dimensions rather than mandatory agent slots. Follow references/review-depth-and-delegation.md: explicit heavy review defaults to at most one active review subagent per review round, while the main agent owns scope, guidance discovery, validation, convergence, and reporting. A review loop may use multiple reviewers sequentially, but each new issue-discovery round must start with a fresh reviewer. Use parallel reviewers within one round only when the user explicitly requests them or a stated high-risk scope justifies independent review.

---name: codereview-familydescription: Structured code review family for explicit heavy review requests and architecture-sensitive review/evaluation requests. Route to PR/MR review, local changeset review, or architecture review based on the review target and intent. Do not use for routine self-checks or casual bug scans.---Use this skill family when the user explicitly asks for a heavy review flow, or when the task is already a review/evaluation request and the auto-trigger rules in `references/triggering-and-signal.md` select a heavy review path.Read `references/triggering-and-signal.md` and`references/review-depth-and-delegation.md` first. When preparing a finalreport, also use `references/report-composition.md`.Then route to the smallest sub-skill that matches the target:- `changeset-codereview/` Use for explicit review of a commit, branch diff, staged diff, working tree diff, patch, or local file set.- `prmr-codereview/` Use for explicit GitHub PR or GitLab MR review when PR/MR-specific transport, metadata, or comment posting is required.- `architecture-review/` Use for explicit architecture review or for review/evaluation requests where layering, contracts, compatibility surfaces, abstractions, or public API design are the main question.If the target is a PR/MR or local changeset and the user explicitly asks forarchitecture review in addition to ordinary review dimensions, or if thereviewed scope materially changes layering, ownership boundaries, dependencydirection, compatibility/public surfaces, or other architecture-reviewdimensions, keep `prmr-codereview` or `changeset-codereview` as the activeskill. For PR/MR reviews, the delegated `changeset-codereview` workflow ownsapplying any paired `architecture-review/` dimensions. A paired skill is alogical report input and normally shares the same review subagent. In bothcases, the active skill owns the main report and must summarize the pairedarchitecture results in `Related review inputs`.Do not use this family for:- routine implementation self-checks- quick bug scans during development- ordinary debugging- style-only or taste-only feedbackUse the smallest review surface that satisfies the request. The heavier theworkflow, the more explicit the user request should be.Treat reviewer roles as logical review dimensions rather than mandatory agentslots. Follow `references/review-depth-and-delegation.md`: explicit heavyreview defaults to at most one active review subagent per review round, whilethe main agent owns scope, guidance discovery, validation, convergence, andreporting. A review loop may use multiple reviewers sequentially, but each newissue-discovery round must start with a fresh reviewer. Use parallel reviewerswithin one round only when the user explicitly requests them or a statedhigh-risk scope justifies independent review.

Frequently asked questions