skills.rest
lfuuu
lfuuuCOMMUNITY
Explore all skills by this author →
GitHub

reviewer-core

Reviews code diffs against plans and returns binary APPROVED or CHANGES_REQUESTED verdicts.

Updated Jun 2, 2026
Search Tags:#code review#checklist#agent workflow#quality gate#git diff#plan-review#verdict
One-click install
npx skills add https://github.com/lfuuu/claude-rules --skill reviewer-core-lfuuu
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: reviewer-core
Source: https://github.com/lfuuu/claude-rules/tree/main/global-skills/reviewer-core
Command: npx skills add https://github.com/lfuuu/claude-rules --skill reviewer-core-lfuuu

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? Code review in agent-driven workflows often drifts: reviewers fix code themselves, run tests they should not, or return vague verdicts like "almost ready". This Skill defines a strict, read-only reviewer role that checks implementation against the assigned plan and returns an unambiguous binary verdict. ## Core Features & Use Cases - Binary verdict contract: Every review ends in APPROVED or CHANGES_REQUESTED, with BLOCK, WARNING, and NOTE findings all blocking the pipeline by default. - Fixed 8-step pipeline: Scope detection, read-only lint/typecheck runs, diff classification, completeness check against the assigned task, checklist application, and a structured report with file:line diagnostics and fix recipes. - Plan-review gate: A dedicated mode compares git history against the plan file to catch silent failures where local checks pass but the overall plan is incomplete. - Use Case: A teamlead agent finishes an implementation cycle and invokes the reviewer; the reviewer greps its runbooks, verifies every assigned task item is implemented, flags a hardcoded secret as BLOCK, and returns CHANGES_REQUESTED with concrete fix instructions. ## Quick Start Import this core from a project wrapper by adding the line "Действует reviewer-core" to the wrapper's SKILL.md, then invoke /reviewer auto to review the latest commit diff.

Frequently Asked Questions about reviewer-core

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I run an automated code review on the latest git commit?▼

Invoke the reviewer in auto mode, which derives the file scope from git diff HEAD~1 or the staged diff. It then runs read-only lint and typecheck commands, classifies findings as BLOCK, WARNING, or NOTE, and returns a single verdict.

What is the difference between BLOCK, WARNING, and NOTE review findings?▼

BLOCK covers architecture, security, and test violations; WARNING covers serious quality and convention issues; NOTE covers minor improvements. By default all three categories block the merge and produce CHANGES_REQUESTED, though a project wrapper may relax minor-only cases.

Can the reviewer skill fix the code issues it finds?▼

No. The reviewer is strictly read-only: it cannot edit code or documents, run tests, commit, or spawn implementer agents. It reports each finding with file, line, rationale, and a fix recipe, and a separate implementer agent applies the fixes.

How does plan-review mode catch incomplete implementations?▼

Plan-review compares the git log since a baseline against the plan file and its artifacts, verifying every assigned task item is implemented or explicitly deferred. Missing items without a deferral note are flagged as BLOCK, catching silent failures local checks miss.

What can a project wrapper override in reviewer-core?▼

A wrapper may define domain BLOCK criteria, review iteration limits, minor-finding leniency, plan-review checklists, mechanical check scripts, and project-specific tool sets. It cannot override the read-only constraint, the no-test rule, the subagent nesting limit, or the commit format rules.

Related Skills

Explore More Software Engineering Skills→
voidcorp-corevoidcorp-core

void-verify

Runs a twelve-item observed pre-flight checklist before reporting any task as done.

Official
Intermediate
0xdeafcafe0xdeafcafe

drive-change

Audit code changes through parallel reviews and apply fixes until findings stabilize.

Community
Advanced
netbox-communitynetbox-community
21.3k

run-tests

Run NetBox Django tests via manage.py with testing configuration.

Official
Intermediate
carlos-algmscarlos-algms
2

orchestrate-agent-clis

Orchestrate multiple AI agent CLIs to compare solution approaches.

Community
Intermediate